[dali_2.3.23] Merge branch 'devel/master'
[platform/core/uifw/dali-toolkit.git] / dali-scene3d / public-api / loader / mesh-definition.cpp
index 10adc2d..6728790 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2023 Samsung Electronics Co., Ltd.
+ * Copyright (c) 2024 Samsung Electronics Co., Ltd.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
  *
  */
 
-// INTERNAL INCLUDES
-#include "dali-scene3d/public-api/loader/mesh-definition.h"
+// CLASS HEADER
+#include <dali-scene3d/public-api/loader/mesh-definition.h>
 
 // EXTERNAL INCLUDES
 #include <dali/devel-api/adaptor-framework/file-stream.h>
+#include <dali/devel-api/adaptor-framework/pixel-buffer.h>
 #include <dali/integration-api/debug.h>
+#include <dali/public-api/math/compile-time-math.h>
 #include <cstring>
 #include <fstream>
-#include "dali/devel-api/adaptor-framework/pixel-buffer.h"
+#include <functional>
+#include <type_traits>
 
-namespace Dali
+namespace Dali::Scene3D::Loader
 {
-namespace Scene3D
+namespace
 {
-namespace Loader
+enum class LoadDataType
 {
-namespace
+  UNSIGNED_BYTE = 0,
+  UNSIGNED_SHORT,
+  FLOAT
+};
+
+struct LoadAccessorInputs
 {
+  MeshDefinition::RawData&  rawData;
+  MeshDefinition::Accessor& accessor;
+  uint32_t                  flags;
+  std::fstream&             fileStream;
+  std::string&              meshPath;
+  BufferDefinition::Vector& buffers;
+};
+
+struct LoadAccessorListInputs
+{
+  MeshDefinition::RawData&               rawData;
+  std::vector<MeshDefinition::Accessor>& accessors;
+  uint32_t                               flags;
+  std::fstream&                          fileStream;
+  std::string&                           meshPath;
+  BufferDefinition::Vector&              buffers;
+};
+
+template<bool use32BitIndices>
 class IndexProvider
 {
 public:
+  using IndexType = typename std::conditional_t<use32BitIndices, uint32_t, uint16_t>;
   IndexProvider(const uint16_t* indices)
   : mData(reinterpret_cast<uintptr_t>(indices)),
     mFunc(indices ? IncrementPointer : Increment)
   {
   }
 
-  uint16_t operator()()
+  IndexType operator()()
   {
     return mFunc(mData);
   }
 
 private:
-  static uint16_t Increment(uintptr_t& data)
+  static IndexType Increment(uintptr_t& data)
   {
-    return static_cast<uint16_t>(data++);
+    // mData was 'zero' at construct time. Just simply return counter start with 0.
+    return static_cast<IndexType>(data++);
   }
 
-  static uint16_t IncrementPointer(uintptr_t& data)
+  static IndexType IncrementPointer(uintptr_t& data)
   {
-    auto iPtr   = reinterpret_cast<const uint16_t*>(data);
+    auto iPtr   = reinterpret_cast<const IndexType*>(data);
     auto result = *iPtr;
     data        = reinterpret_cast<uintptr_t>(++iPtr);
     return result;
   }
 
   uintptr_t mData;
-  uint16_t (*mFunc)(uintptr_t&);
+  IndexType (*mFunc)(uintptr_t&);
 };
 
-const std::string QUAD("quad");
+const char* QUAD("quad");
 
 ///@brief Reads a blob from the given stream @a source into @a target, which must have
 /// at least @a descriptor.length bytes.
 bool ReadBlob(const MeshDefinition::Blob& descriptor, std::istream& source, uint8_t* target)
 {
   source.clear();
-  if(!source.seekg(descriptor.mOffset, std::istream::beg))
+  if(!source.seekg(static_cast<std::streamoff>(static_cast<std::size_t>(descriptor.mOffset)), std::istream::beg))
   {
     return false;
   }
 
   if(descriptor.IsConsecutive())
   {
-    return !!source.read(reinterpret_cast<char*>(target), descriptor.mLength);
+    return !!source.read(reinterpret_cast<char*>(target), static_cast<std::streamsize>(static_cast<size_t>(descriptor.mLength)));
   }
   else
   {
@@ -89,11 +118,11 @@ bool ReadBlob(const MeshDefinition::Blob& descriptor, std::istream& source, uint
       uint32_t       readSize  = 0;
       uint32_t       totalSize = (descriptor.mLength / descriptor.mElementSizeHint) * descriptor.mStride;
       while(readSize < totalSize &&
-            source.read(reinterpret_cast<char*>(target), descriptor.mElementSizeHint) &&
-            source.seekg(diff, std::istream::cur))
+            source.read(reinterpret_cast<char*>(target), descriptor.mElementSizeHint))
       {
         readSize += descriptor.mStride;
         target += descriptor.mElementSizeHint;
+        source.seekg(static_cast<std::streamoff>(static_cast<std::size_t>(diff)), std::istream::cur);
       }
       return readSize == totalSize;
     }
@@ -112,7 +141,7 @@ void ReadValues(const std::vector<uint8_t>& valuesBuffer, const std::vector<uint
   }
 }
 
-bool ReadAccessor(const MeshDefinition::Accessor& accessor, std::istream& source, uint8_t* target)
+bool ReadAccessor(const MeshDefinition::Accessor& accessor, std::istream& source, uint8_t* target, std::vector<uint32_t>* sparseIndices)
 {
   bool success = false;
 
@@ -151,72 +180,136 @@ bool ReadAccessor(const MeshDefinition::Accessor& accessor, std::istream& source
       return false;
     }
 
+    // If non-null sparse indices vector, prepare it for output
+    if(sparseIndices)
+    {
+      sparseIndices->resize(accessor.mSparse->mCount);
+    }
+
     switch(indices.mElementSizeHint)
     {
       case 1u:
       {
         ReadValues<uint8_t>(valuesBuffer, indicesBuffer, target, accessor.mSparse->mCount, values.mElementSizeHint);
+        if(sparseIndices)
+        {
+          // convert 8-bit indices into 32-bit
+          std::transform(indicesBuffer.begin(), indicesBuffer.end(), sparseIndices->begin(), [](const uint8_t& value) { return uint32_t(value); });
+        }
         break;
       }
       case 2u:
       {
         ReadValues<uint16_t>(valuesBuffer, indicesBuffer, target, accessor.mSparse->mCount, values.mElementSizeHint);
+        if(sparseIndices)
+        {
+          // convert 16-bit indices into 32-bit
+          std::transform(reinterpret_cast<uint16_t*>(indicesBuffer.data()),
+                         reinterpret_cast<uint16_t*>(indicesBuffer.data()) + accessor.mSparse->mCount,
+                         sparseIndices->begin(),
+                         [](const uint16_t& value) {
+                           return uint32_t(value);
+                         });
+        }
         break;
       }
       case 4u:
       {
         ReadValues<uint32_t>(valuesBuffer, indicesBuffer, target, accessor.mSparse->mCount, values.mElementSizeHint);
+        if(sparseIndices)
+        {
+          std::copy(indicesBuffer.begin(), indicesBuffer.end(), reinterpret_cast<uint8_t*>(sparseIndices->data()));
+        }
         break;
       }
       default:
+      {
         DALI_ASSERT_DEBUG(!"Unsupported type for an index");
+      }
     }
   }
 
   return success;
 }
 
-template<typename T>
-void ReadJointAccessor(MeshDefinition::RawData& raw, const MeshDefinition::Accessor& accessor, std::istream& source, const std::string& meshPath)
+bool ReadAccessor(const MeshDefinition::Accessor& accessor, std::istream& source, uint8_t* target)
+{
+  return ReadAccessor(accessor, source, target, nullptr);
+}
+
+template<typename T, bool needsNormalize>
+void ReadVectorAccessor(const MeshDefinition::Accessor& accessor, std::istream& source, std::vector<uint8_t>& buffer)
 {
   constexpr auto sizeofBlobUnit = sizeof(T) * 4;
 
   DALI_ASSERT_ALWAYS(((accessor.mBlob.mLength % sizeofBlobUnit == 0) ||
                       accessor.mBlob.mStride >= sizeofBlobUnit) &&
-                     "Joints buffer length not a multiple of element size");
+                     "Buffer length not a multiple of element size");
   const auto inBufferSize  = accessor.mBlob.GetBufferSize();
   const auto outBufferSize = (sizeof(Vector4) / sizeofBlobUnit) * inBufferSize;
 
-  std::vector<uint8_t> buffer(outBufferSize);
-  auto                 inBuffer = buffer.data() + outBufferSize - inBufferSize;
+  buffer.resize(outBufferSize);
+  auto inBuffer = buffer.data() + outBufferSize - inBufferSize;
   if(!ReadAccessor(accessor, source, inBuffer))
   {
-    ExceptionFlinger(ASSERT_LOCATION) << "Failed to read joints from '" << meshPath << "'.";
+    ExceptionFlinger(ASSERT_LOCATION) << "Failed to read vector data from Accessor.";
   }
 
-  if constexpr(sizeofBlobUnit != sizeof(Vector4))
+  if(sizeofBlobUnit != sizeof(Vector4))
   {
     auto       floats = reinterpret_cast<float*>(buffer.data());
     const auto end    = inBuffer + inBufferSize;
     while(inBuffer != end)
     {
       const auto value = *reinterpret_cast<T*>(inBuffer);
-      *floats          = static_cast<float>(value);
+      *floats          = (needsNormalize) ? static_cast<float>(value) / static_cast<float>((1 << (sizeof(T) * 8)) - 1) : static_cast<float>(value);
 
       inBuffer += sizeof(T);
       ++floats;
     }
   }
-  raw.mAttribs.push_back({"aJoints", Property::VECTOR4, static_cast<uint32_t>(outBufferSize / sizeof(Vector4)), std::move(buffer)});
 }
 
-void GenerateNormals(MeshDefinition::RawData& raw)
+template<bool needsNormalize>
+void ReadTypedVectorAccessor(LoadDataType loadDataType, MeshDefinition::Accessor& accessor, std::iostream& stream, std::vector<uint8_t>& buffer)
+{
+  switch(loadDataType)
+  {
+    case LoadDataType::UNSIGNED_SHORT:
+    {
+      ReadVectorAccessor<uint16_t, needsNormalize>(accessor, stream, buffer);
+      break;
+    }
+    case LoadDataType::UNSIGNED_BYTE:
+    {
+      ReadVectorAccessor<uint8_t, needsNormalize>(accessor, stream, buffer);
+      break;
+    }
+    default:
+    {
+      ReadVectorAccessor<float, needsNormalize>(accessor, stream, buffer);
+      break;
+    }
+  }
+}
+
+template<bool use32BitsIndices, typename IndexProviderType = IndexProvider<use32BitsIndices>>
+bool GenerateNormals(MeshDefinition::RawData& raw)
 {
+  using IndexType = typename IndexProviderType::IndexType;
+
+  // mIndicies size must be even if we use 32bit indices.
+  if(DALI_UNLIKELY(use32BitsIndices && !raw.mIndices.empty() && !(raw.mIndices.size() % (sizeof(IndexType) / sizeof(uint16_t)) == 0)))
+  {
+    return false;
+  }
+
   auto& attribs = raw.mAttribs;
   DALI_ASSERT_DEBUG(attribs.size() > 0); // positions
-  IndexProvider getIndex(raw.mIndices.data());
 
-  const uint32_t numIndices = raw.mIndices.empty() ? attribs[0].mNumElements : static_cast<uint32_t>(raw.mIndices.size());
+  IndexProviderType getIndex(raw.mIndices.data());
+
+  const uint32_t numIndices = raw.mIndices.empty() ? attribs[0].mNumElements : static_cast<uint32_t>(raw.mIndices.size() / (sizeof(IndexType) / sizeof(uint16_t)));
 
   auto* positions = reinterpret_cast<const Vector3*>(attribs[0].mData.data());
 
@@ -225,8 +318,8 @@ void GenerateNormals(MeshDefinition::RawData& raw)
 
   for(uint32_t i = 0; i < numIndices; i += 3)
   {
-    uint16_t indices[]{getIndex(), getIndex(), getIndex()};
-    Vector3  pos[]{positions[indices[0]], positions[indices[1]], positions[indices[2]]};
+    IndexType indices[]{getIndex(), getIndex(), getIndex()};
+    Vector3   pos[]{positions[indices[0]], positions[indices[1]], positions[indices[2]]};
 
     Vector3 a = pos[1] - pos[0];
     Vector3 b = pos[2] - pos[0];
@@ -245,85 +338,103 @@ void GenerateNormals(MeshDefinition::RawData& raw)
   }
 
   attribs.push_back({"aNormal", Property::VECTOR3, attribs[0].mNumElements, std::move(buffer)});
+
+  return true;
 }
 
-void GenerateTangentsWithUvs(MeshDefinition::RawData& raw)
+template<bool use32BitsIndices, bool useVec3, bool hasUvs, typename T = std::conditional_t<useVec3, Vector3, Vector4>, typename = std::enable_if_t<(std::is_same<T, Vector3>::value || std::is_same<T, Vector4>::value)>, typename IndexProviderType = IndexProvider<use32BitsIndices>>
+bool GenerateTangents(MeshDefinition::RawData& raw)
 {
-  auto& attribs = raw.mAttribs;
-  DALI_ASSERT_DEBUG(attribs.size() > 2); // positions, normals, uvs
-  IndexProvider getIndex(raw.mIndices.data());
-
-  const uint32_t numIndices = raw.mIndices.empty() ? attribs[0].mNumElements : static_cast<uint32_t>(raw.mIndices.size());
-
-  auto* positions = reinterpret_cast<const Vector3*>(attribs[0].mData.data());
-  auto* uvs       = reinterpret_cast<const Vector2*>(attribs[2].mData.data());
+  using IndexType = typename IndexProviderType::IndexType;
 
-  std::vector<uint8_t> buffer(attribs[0].mNumElements * sizeof(Vector3));
-  auto                 tangents = reinterpret_cast<Vector3*>(buffer.data());
+  // mIndicies size must be even if we use 32bit indices.
+  if(DALI_UNLIKELY(use32BitsIndices && !raw.mIndices.empty() && !(raw.mIndices.size() % (sizeof(IndexType) / sizeof(uint16_t)) == 0)))
+  {
+    return false;
+  }
 
-  for(uint32_t i = 0; i < numIndices; i += 3)
+  auto& attribs = raw.mAttribs;
+  // Required positions, normals, uvs (if we have). If not, skip generation
+  if(DALI_UNLIKELY(attribs.size() < (2 + static_cast<size_t>(hasUvs))))
   {
-    uint16_t indices[]{getIndex(), getIndex(), getIndex()};
-    Vector3  pos[]{positions[indices[0]], positions[indices[1]], positions[indices[2]]};
-    Vector2  uv[]{uvs[indices[0]], uvs[indices[1]], uvs[indices[2]]};
+    return false;
+  }
 
-    float x0 = pos[1].x - pos[0].x;
-    float y0 = pos[1].y - pos[0].y;
-    float z0 = pos[1].z - pos[0].z;
+  std::vector<uint8_t> buffer(attribs[0].mNumElements * sizeof(T));
+  auto                 tangents = reinterpret_cast<T*>(buffer.data());
 
-    float x1 = pos[2].x - pos[0].x;
-    float y1 = pos[2].y - pos[0].y;
-    float z1 = pos[2].z - pos[0].z;
+  if constexpr(hasUvs)
+  {
+    IndexProviderType getIndex(raw.mIndices.data());
 
-    float s0 = uv[1].x - uv[0].x;
-    float t0 = uv[1].y - uv[0].y;
+    const uint32_t numIndices = raw.mIndices.empty() ? attribs[0].mNumElements : static_cast<uint32_t>(raw.mIndices.size() / (sizeof(IndexType) / sizeof(uint16_t)));
 
-    float s1 = uv[2].x - uv[0].x;
-    float t1 = uv[2].y - uv[0].y;
+    auto* positions = reinterpret_cast<const Vector3*>(attribs[0].mData.data());
+    auto* uvs       = reinterpret_cast<const Vector2*>(attribs[2].mData.data());
 
-    float   r = 1.f / (s0 * t1 - t0 * s1);
-    Vector3 tangent((x0 * t1 - t0 * x1) * r, (y0 * t1 - t0 * y1) * r, (z0 * t1 - t0 * z1) * r);
-    tangents[indices[0]] += tangent;
-    tangents[indices[1]] += tangent;
-    tangents[indices[2]] += tangent;
+    for(uint32_t i = 0; i < numIndices; i += 3)
+    {
+      IndexType indices[]{getIndex(), getIndex(), getIndex()};
+      Vector3   pos[]{positions[indices[0]], positions[indices[1]], positions[indices[2]]};
+      Vector2   uv[]{uvs[indices[0]], uvs[indices[1]], uvs[indices[2]]};
+
+      float x0 = pos[1].x - pos[0].x;
+      float y0 = pos[1].y - pos[0].y;
+      float z0 = pos[1].z - pos[0].z;
+
+      float x1 = pos[2].x - pos[0].x;
+      float y1 = pos[2].y - pos[0].y;
+      float z1 = pos[2].z - pos[0].z;
+
+      float s0 = uv[1].x - uv[0].x;
+      float t0 = uv[1].y - uv[0].y;
+
+      float s1 = uv[2].x - uv[0].x;
+      float t1 = uv[2].y - uv[0].y;
+
+      float   det = (s0 * t1 - t0 * s1);
+      float   r   = 1.f / ((std::abs(det) < Dali::Epsilon<1000>::value) ? (Dali::Epsilon<1000>::value * (det > 0.0f ? 1.f : -1.f)) : det);
+      Vector3 tangent((x0 * t1 - t0 * x1) * r, (y0 * t1 - t0 * y1) * r, (z0 * t1 - t0 * z1) * r);
+      tangents[indices[0]] += T(tangent);
+      tangents[indices[1]] += T(tangent);
+      tangents[indices[2]] += T(tangent);
+    }
   }
 
   auto* normals = reinterpret_cast<const Vector3*>(attribs[1].mData.data());
   auto  iEnd    = normals + attribs[1].mNumElements;
   while(normals != iEnd)
   {
-    *tangents -= *normals * normals->Dot(*tangents);
-    tangents->Normalize();
-
-    ++tangents;
-    ++normals;
-  }
-  attribs.push_back({"aTangent", Property::VECTOR3, attribs[0].mNumElements, std::move(buffer)});
-}
-
-void GenerateTangents(MeshDefinition::RawData& raw)
-{
-  auto& attribs = raw.mAttribs;
-  DALI_ASSERT_DEBUG(attribs.size() > 1); // positions, normals
-
-  auto* normals = reinterpret_cast<const Vector3*>(attribs[1].mData.data());
-
-  std::vector<uint8_t> buffer(attribs[0].mNumElements * sizeof(Vector3));
-  auto                 tangents = reinterpret_cast<Vector3*>(buffer.data());
-
-  auto iEnd = normals + attribs[1].mNumElements;
-  while(normals != iEnd)
-  {
-    Vector3 t[]{normals->Cross(Vector3::XAXIS), normals->Cross(Vector3::YAXIS)};
+    Vector3 tangentVec3;
+    if constexpr(hasUvs)
+    {
+      // Calculated by indexs
+      tangentVec3 = Vector3((*tangents).x, (*tangents).y, (*tangents).z);
+    }
+    else
+    {
+      // Only choiced by normal vector. by indexs
+      Vector3 t[]{normals->Cross(Vector3::XAXIS), normals->Cross(Vector3::YAXIS)};
+      tangentVec3 = t[t[1].LengthSquared() > t[0].LengthSquared()];
+    }
 
-    *tangents = t[t[1].LengthSquared() > t[0].LengthSquared()];
-    *tangents -= *normals * normals->Dot(*tangents);
-    tangents->Normalize();
+    tangentVec3 -= *normals * normals->Dot(tangentVec3);
+    tangentVec3.Normalize();
+    if constexpr(useVec3)
+    {
+      *tangents = tangentVec3;
+    }
+    else
+    {
+      *tangents = Vector4(tangentVec3.x, tangentVec3.y, tangentVec3.z, 1.0f);
+    }
 
     ++tangents;
     ++normals;
   }
-  attribs.push_back({"aTangent", Property::VECTOR3, attribs[0].mNumElements, std::move(buffer)});
+  attribs.push_back({"aTangent", useVec3 ? Property::VECTOR3 : Property::VECTOR4, attribs[0].mNumElements, std::move(buffer)});
+
+  return true;
 }
 
 void CalculateTextureSize(uint32_t totalTextureSize, uint32_t& textureWidth, uint32_t& textureHeight)
@@ -350,58 +461,221 @@ void CalculateTextureSize(uint32_t totalTextureSize, uint32_t& textureWidth, uin
   textureHeight = 1u << powHeight;
 }
 
-void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, const std::vector<MeshDefinition::BlendShape>& blendShapes, uint32_t numberOfVertices, float& blendShapeUnnormalizeFactor, BufferDefinition::Vector& buffers)
+template<typename T>
+float GetNormalizedScale()
+{
+  return 1.0f / (std::numeric_limits<T>::max());
+}
+
+template<typename T>
+void DequantizeData(std::vector<uint8_t>& buffer, float* dequantizedValues, uint32_t numValues, bool normalized)
+{
+  // see https://github.com/KhronosGroup/glTF/tree/master/extensions/2.0/Khronos/KHR_mesh_quantization#encoding-quantized-data
+
+  T* values = reinterpret_cast<T*>(buffer.data());
+
+  for(uint32_t i = 0; i < numValues; ++i)
+  {
+    *dequantizedValues = normalized ? std::max((*values) * GetNormalizedScale<T>(), -1.0f) : *values;
+
+    values++;
+    dequantizedValues++;
+  }
+}
+
+void GetDequantizedData(std::vector<uint8_t>& buffer, uint32_t numComponents, uint32_t count, uint32_t flags, bool normalized)
+{
+  bool dequantized = false;
+
+  std::vector<uint8_t> dequantizedBuffer(count * numComponents * sizeof(float));
+  float*               dequantizedValues = reinterpret_cast<float*>(dequantizedBuffer.data());
+
+  if(MaskMatch(flags, MeshDefinition::Flags::S8_POSITION) || MaskMatch(flags, MeshDefinition::Flags::S8_NORMAL) || MaskMatch(flags, MeshDefinition::Flags::S8_TANGENT) || MaskMatch(flags, MeshDefinition::Flags::S8_TEXCOORD))
+  {
+    DequantizeData<int8_t>(buffer, dequantizedValues, numComponents * count, normalized);
+    dequantized = true;
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::U8_POSITION) || MaskMatch(flags, MeshDefinition::Flags::U8_TEXCOORD))
+  {
+    DequantizeData<uint8_t>(buffer, dequantizedValues, numComponents * count, normalized);
+    dequantized = true;
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::S16_POSITION) || MaskMatch(flags, MeshDefinition::Flags::S16_NORMAL) || MaskMatch(flags, MeshDefinition::Flags::S16_TANGENT) || MaskMatch(flags, MeshDefinition::Flags::S16_TEXCOORD))
+  {
+    DequantizeData<int16_t>(buffer, dequantizedValues, numComponents * count, normalized);
+    dequantized = true;
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::U16_POSITION) || MaskMatch(flags, MeshDefinition::Flags::U16_TEXCOORD))
+  {
+    DequantizeData<uint16_t>(buffer, dequantizedValues, numComponents * count, normalized);
+    dequantized = true;
+  }
+
+  if(dequantized)
+  {
+    buffer = std::move(dequantizedBuffer);
+  }
+}
+
+void GetDequantizedMinMax(std::vector<float>& min, std::vector<float>& max, uint32_t flags)
+{
+  float scale = 1.0f;
+
+  if(MaskMatch(flags, MeshDefinition::Flags::S8_POSITION) || MaskMatch(flags, MeshDefinition::Flags::S8_NORMAL) || MaskMatch(flags, MeshDefinition::Flags::S8_TANGENT) || MaskMatch(flags, MeshDefinition::Flags::S8_TEXCOORD))
+  {
+    scale = GetNormalizedScale<int8_t>();
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::U8_POSITION) || MaskMatch(flags, MeshDefinition::Flags::U8_TEXCOORD))
+  {
+    scale = GetNormalizedScale<uint8_t>();
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::S16_POSITION) || MaskMatch(flags, MeshDefinition::Flags::S16_NORMAL) || MaskMatch(flags, MeshDefinition::Flags::S16_TANGENT) || MaskMatch(flags, MeshDefinition::Flags::S16_TEXCOORD))
+  {
+    scale = GetNormalizedScale<int16_t>();
+  }
+  else if(MaskMatch(flags, MeshDefinition::Flags::U16_POSITION) || MaskMatch(flags, MeshDefinition::Flags::U16_TEXCOORD))
+  {
+    scale = GetNormalizedScale<uint16_t>();
+  }
+
+  if(scale != 1.0f)
+  {
+    for(float& value : min)
+    {
+      value = std::max(value * scale, -1.0f);
+    }
+
+    for(float& value : max)
+    {
+      value = std::min(value * scale, 1.0f);
+    }
+  }
+}
+
+void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, std::vector<MeshDefinition::BlendShape>& blendShapes, uint32_t numberOfVertices, float& blendShapeUnnormalizeFactor, BufferDefinition::Vector& buffers)
 {
   uint32_t geometryBufferIndex = 0u;
-  float    maxDistance         = 0.f;
+  float    maxDistanceSquared  = 0.f;
   Vector3* geometryBufferV3    = reinterpret_cast<Vector3*>(geometryBuffer);
-  for(const auto& blendShape : blendShapes)
+  for(auto& blendShape : blendShapes)
   {
     if(blendShape.deltas.IsDefined())
     {
-      DALI_ASSERT_ALWAYS(((blendShape.deltas.mBlob.mLength % sizeof(Vector3) == 0u) ||
-                          blendShape.deltas.mBlob.mStride >= sizeof(Vector3)) &&
-                         "Blend Shape position buffer length not a multiple of element size");
+      const auto bufferSize = blendShape.deltas.mBlob.GetBufferSize();
+      uint32_t   numVector3;
 
-      const auto           bufferSize = blendShape.deltas.mBlob.GetBufferSize();
-      std::vector<uint8_t> buffer(bufferSize);
-      if(ReadAccessor(blendShape.deltas, buffers[blendShape.deltas.mBufferIdx].GetBufferStream(), buffer.data()))
+      if(MaskMatch(blendShape.mFlags, MeshDefinition::S8_POSITION))
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.deltas.mBlob.mLength % (sizeof(uint8_t) * 3) == 0) ||
+                            blendShape.deltas.mBlob.mStride >= (sizeof(uint8_t) * 3)) &&
+                           "Blend Shape position buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(uint8_t) * 3));
+      }
+      else if(MaskMatch(blendShape.mFlags, MeshDefinition::S16_POSITION))
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.deltas.mBlob.mLength % (sizeof(uint16_t) * 3) == 0) ||
+                            blendShape.deltas.mBlob.mStride >= (sizeof(uint16_t) * 3)) &&
+                           "Blend Shape position buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(uint16_t) * 3));
+      }
+      else
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.deltas.mBlob.mLength % sizeof(Vector3) == 0) ||
+                            blendShape.deltas.mBlob.mStride >= sizeof(Vector3)) &&
+                           "Blend Shape position buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
+      }
+
+      std::vector<uint8_t>  buffer(bufferSize);
+      std::vector<uint32_t> sparseIndices{};
+
+      if(ReadAccessor(blendShape.deltas, buffers[blendShape.deltas.mBufferIdx].GetBufferStream(), buffer.data(), &sparseIndices))
       {
-        blendShape.deltas.mBlob.ApplyMinMax(static_cast<uint32_t>(bufferSize / sizeof(Vector3)), reinterpret_cast<float*>(buffer.data()));
+        GetDequantizedData(buffer, 3u, numVector3, blendShape.mFlags & MeshDefinition::POSITIONS_MASK, blendShape.deltas.mNormalized);
+
+        if(blendShape.deltas.mNormalized)
+        {
+          GetDequantizedMinMax(blendShape.deltas.mBlob.mMin, blendShape.deltas.mBlob.mMax, blendShape.mFlags & MeshDefinition::POSITIONS_MASK);
+        }
+
+        blendShape.deltas.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()), &sparseIndices);
+
         // Calculate the difference with the original mesh.
         // Find the max distance to normalize the deltas.
-        const Vector3* const deltasBuffer = reinterpret_cast<const Vector3* const>(buffer.data());
+        const auto* const deltasBuffer = reinterpret_cast<const Vector3* const>(buffer.data());
 
-        for(uint32_t index = 0u; index < numberOfVertices; ++index)
-        {
-          Vector3& delta = geometryBufferV3[geometryBufferIndex++];
-          delta          = deltasBuffer[index];
+        auto ProcessVertex = [&geometryBufferV3, &deltasBuffer, &maxDistanceSquared](uint32_t geometryBufferIndex, uint32_t deltaIndex) {
+          Vector3& delta = geometryBufferV3[geometryBufferIndex] = deltasBuffer[deltaIndex];
+          delta                                                  = deltasBuffer[deltaIndex];
+          return std::max(maxDistanceSquared, delta.LengthSquared());
+        };
 
-          maxDistance = std::max(maxDistance, delta.LengthSquared());
+        if(sparseIndices.empty())
+        {
+          for(uint32_t index = 0u; index < numberOfVertices; ++index)
+          {
+            maxDistanceSquared = ProcessVertex(geometryBufferIndex++, index);
+          }
+        }
+        else
+        {
+          // initialize blendshape texture
+          // TODO: there may be a case when sparse accessor uses a base buffer view for initial values.
+          std::fill(geometryBufferV3 + geometryBufferIndex, geometryBufferV3 + geometryBufferIndex + numberOfVertices, Vector3::ZERO);
+          for(auto index : sparseIndices)
+          {
+            maxDistanceSquared = ProcessVertex(geometryBufferIndex + index, index);
+          }
+          geometryBufferIndex += numberOfVertices;
         }
       }
     }
 
     if(blendShape.normals.IsDefined())
     {
-      DALI_ASSERT_ALWAYS(((blendShape.normals.mBlob.mLength % sizeof(Vector3) == 0u) ||
-                          blendShape.normals.mBlob.mStride >= sizeof(Vector3)) &&
-                         "Blend Shape normals buffer length not a multiple of element size");
+      const auto bufferSize = blendShape.normals.mBlob.GetBufferSize();
+      uint32_t   numVector3;
 
-      const auto           bufferSize = blendShape.normals.mBlob.GetBufferSize();
-      std::vector<uint8_t> buffer(bufferSize);
-      if(ReadAccessor(blendShape.normals, buffers[blendShape.normals.mBufferIdx].GetBufferStream(), buffer.data()))
+      if(MaskMatch(blendShape.mFlags, MeshDefinition::S8_NORMAL))
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.normals.mBlob.mLength % (sizeof(int8_t) * 3) == 0) ||
+                            blendShape.normals.mBlob.mStride >= (sizeof(int8_t) * 3)) &&
+                           "Blend Shape normals buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int8_t) * 3));
+      }
+      else if(MaskMatch(blendShape.mFlags, MeshDefinition::S16_NORMAL))
       {
-        blendShape.normals.mBlob.ApplyMinMax(static_cast<uint32_t>(bufferSize / sizeof(Vector3)), reinterpret_cast<float*>(buffer.data()));
+        DALI_ASSERT_ALWAYS(((blendShape.normals.mBlob.mLength % (sizeof(int16_t) * 3) == 0) ||
+                            blendShape.normals.mBlob.mStride >= (sizeof(int16_t) * 3)) &&
+                           "Blend Shape normals buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int16_t) * 3));
+      }
+      else
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.normals.mBlob.mLength % sizeof(Vector3) == 0) ||
+                            blendShape.normals.mBlob.mStride >= sizeof(Vector3)) &&
+                           "Blend Shape normals buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
+      }
 
-        // Calculate the difference with the original mesh, and translate to make all values positive.
-        const Vector3* const deltasBuffer = reinterpret_cast<const Vector3* const>(buffer.data());
+      std::vector<uint8_t>  buffer(bufferSize);
+      std::vector<uint32_t> sparseIndices;
 
-        for(uint32_t index = 0u; index < numberOfVertices; ++index)
+      if(ReadAccessor(blendShape.normals, buffers[blendShape.normals.mBufferIdx].GetBufferStream(), buffer.data(), &sparseIndices))
+      {
+        GetDequantizedData(buffer, 3u, numVector3, blendShape.mFlags & MeshDefinition::NORMALS_MASK, blendShape.normals.mNormalized);
+
+        if(blendShape.normals.mNormalized)
         {
-          Vector3& delta = geometryBufferV3[geometryBufferIndex++];
-          delta          = deltasBuffer[index];
+          GetDequantizedMinMax(blendShape.normals.mBlob.mMin, blendShape.normals.mBlob.mMax, blendShape.mFlags & MeshDefinition::NORMALS_MASK);
+        }
+
+        blendShape.normals.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()), &sparseIndices);
 
+        // Calculate the difference with the original mesh, and translate to make all values positive.
+        const Vector3* const deltasBuffer  = reinterpret_cast<const Vector3* const>(buffer.data());
+        auto                 ProcessVertex = [&geometryBufferV3, &deltasBuffer, &maxDistanceSquared](uint32_t geometryBufferIndex, uint32_t deltaIndex) {
+          Vector3& delta = geometryBufferV3[geometryBufferIndex] = deltasBuffer[deltaIndex];
           delta.x *= 0.5f;
           delta.y *= 0.5f;
           delta.z *= 0.5f;
@@ -409,30 +683,73 @@ void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, const std::vector<MeshDe
           delta.x += 0.5f;
           delta.y += 0.5f;
           delta.z += 0.5f;
+        };
+
+        if(sparseIndices.empty())
+        {
+          for(uint32_t index = 0u; index < numberOfVertices; ++index)
+          {
+            ProcessVertex(geometryBufferIndex++, index);
+          }
+        }
+        else
+        {
+          std::fill(geometryBufferV3 + geometryBufferIndex, geometryBufferV3 + geometryBufferIndex + numberOfVertices, Vector3(0.5, 0.5, 0.5));
+          for(auto index : sparseIndices)
+          {
+            ProcessVertex(geometryBufferIndex + index, index);
+          }
+          geometryBufferIndex += numberOfVertices;
         }
       }
     }
 
     if(blendShape.tangents.IsDefined())
     {
-      DALI_ASSERT_ALWAYS(((blendShape.tangents.mBlob.mLength % sizeof(Vector3) == 0u) ||
-                          blendShape.tangents.mBlob.mStride >= sizeof(Vector3)) &&
-                         "Blend Shape tangents buffer length not a multiple of element size");
+      const auto bufferSize = blendShape.tangents.mBlob.GetBufferSize();
 
-      const auto           bufferSize = blendShape.tangents.mBlob.GetBufferSize();
-      std::vector<uint8_t> buffer(bufferSize);
-      if(ReadAccessor(blendShape.tangents, buffers[blendShape.tangents.mBufferIdx].GetBufferStream(), buffer.data()))
+      uint32_t numVector3;
+
+      if(MaskMatch(blendShape.mFlags, MeshDefinition::S8_TANGENT))
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.tangents.mBlob.mLength % (sizeof(int8_t) * 3) == 0) ||
+                            blendShape.tangents.mBlob.mStride >= (sizeof(int8_t) * 3)) &&
+                           "Blend Shape tangents buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int8_t) * 3));
+      }
+      else if(MaskMatch(blendShape.mFlags, MeshDefinition::S16_TANGENT))
+      {
+        DALI_ASSERT_ALWAYS(((blendShape.tangents.mBlob.mLength % (sizeof(int16_t) * 3) == 0) ||
+                            blendShape.tangents.mBlob.mStride >= (sizeof(int16_t) * 3)) &&
+                           "Blend Shape tangents buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int16_t) * 3));
+      }
+      else
       {
-        blendShape.tangents.mBlob.ApplyMinMax(static_cast<uint32_t>(bufferSize / sizeof(Vector3)), reinterpret_cast<float*>(buffer.data()));
+        DALI_ASSERT_ALWAYS(((blendShape.tangents.mBlob.mLength % sizeof(Vector3) == 0) ||
+                            blendShape.tangents.mBlob.mStride >= sizeof(Vector3)) &&
+                           "Blend Shape tangents buffer length not a multiple of element size");
+        numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
+      }
 
-        // Calculate the difference with the original mesh, and translate to make all values positive.
-        const Vector3* const deltasBuffer = reinterpret_cast<const Vector3* const>(buffer.data());
+      std::vector<uint8_t>  buffer(bufferSize);
+      std::vector<uint32_t> sparseIndices;
+
+      if(ReadAccessor(blendShape.tangents, buffers[blendShape.tangents.mBufferIdx].GetBufferStream(), buffer.data(), &sparseIndices))
+      {
+        GetDequantizedData(buffer, 3u, numVector3, blendShape.mFlags & MeshDefinition::TANGENTS_MASK, blendShape.tangents.mNormalized);
 
-        for(uint32_t index = 0u; index < numberOfVertices; ++index)
+        if(blendShape.tangents.mNormalized)
         {
-          Vector3& delta = geometryBufferV3[geometryBufferIndex++];
-          delta          = deltasBuffer[index];
+          GetDequantizedMinMax(blendShape.tangents.mBlob.mMin, blendShape.tangents.mBlob.mMax, blendShape.mFlags & MeshDefinition::TANGENTS_MASK);
+        }
 
+        blendShape.tangents.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()), &sparseIndices);
+
+        // Calculate the difference with the original mesh, and translate to make all values positive.
+        const Vector3* const deltasBuffer  = reinterpret_cast<const Vector3* const>(buffer.data());
+        auto                 ProcessVertex = [&geometryBufferV3, &deltasBuffer, &maxDistanceSquared](uint32_t geometryBufferIndex, uint32_t deltaIndex) {
+          Vector3& delta = geometryBufferV3[geometryBufferIndex] = deltasBuffer[deltaIndex];
           delta.x *= 0.5f;
           delta.y *= 0.5f;
           delta.z *= 0.5f;
@@ -440,12 +757,37 @@ void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, const std::vector<MeshDe
           delta.x += 0.5f;
           delta.y += 0.5f;
           delta.z += 0.5f;
+        };
+
+        if(sparseIndices.empty())
+        {
+          for(uint32_t index = 0u; index < numberOfVertices; ++index)
+          {
+            ProcessVertex(geometryBufferIndex++, index);
+          }
+        }
+        else
+        {
+          std::fill(geometryBufferV3 + geometryBufferIndex, geometryBufferV3 + geometryBufferIndex + numberOfVertices, Vector3(0.5, 0.5, 0.5));
+          for(auto index : sparseIndices)
+          {
+            ProcessVertex(geometryBufferIndex + index, index);
+          }
+          geometryBufferIndex += numberOfVertices;
         }
       }
     }
   }
 
   geometryBufferIndex = 0u;
+
+  const float maxDistance = sqrtf(maxDistanceSquared);
+
+  const float normalizeFactor = (maxDistanceSquared < Math::MACHINE_EPSILON_100) ? 1.f : (0.5f / maxDistance);
+
+  // Calculate and store the unnormalize factor.
+  blendShapeUnnormalizeFactor = maxDistance * 2.0f;
+
   for(const auto& blendShape : blendShapes)
   {
     // Normalize all the deltas and translate to a possitive value.
@@ -453,8 +795,6 @@ void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, const std::vector<MeshDe
     // whose values that are less than zero are clamped.
     if(blendShape.deltas.IsDefined())
     {
-      const float normalizeFactor = (fabsf(maxDistance) < Math::MACHINE_EPSILON_1000) ? 1.f : (0.5f / sqrtf(maxDistance));
-
       for(uint32_t index = 0u; index < numberOfVertices; ++index)
       {
         Vector3& delta = geometryBufferV3[geometryBufferIndex++];
@@ -462,9 +802,6 @@ void CalculateGltf2BlendShapes(uint8_t* geometryBuffer, const std::vector<MeshDe
         delta.y        = Clamp(((delta.y * normalizeFactor) + 0.5f), 0.f, 1.f);
         delta.z        = Clamp(((delta.z * normalizeFactor) + 0.5f), 0.f, 1.f);
       }
-
-      // Calculate and store the unnormalize factor.
-      blendShapeUnnormalizeFactor = 1.f / normalizeFactor;
     }
 
     if(blendShape.normals.IsDefined())
@@ -486,320 +823,262 @@ std::iostream& GetAvailableData(std::fstream& meshStream, const std::string& mes
   return stream;
 }
 
-} // namespace
-
-MeshDefinition::SparseBlob::SparseBlob(const Blob& indices, const Blob& values, uint32_t count)
-: mIndices{indices},
-  mValues{values},
-  mCount{count}
+template<bool needsNormalize>
+void ReadTypedVectorAccessors(LoadAccessorListInputs loadAccessorListInputs, LoadDataType loadDataType, std::string attributeName)
 {
+  int setIndex = 0;
+  for(auto& accessor : loadAccessorListInputs.accessors)
+  {
+    std::string        pathJoint;
+    auto&              dataStream = GetAvailableData(loadAccessorListInputs.fileStream, loadAccessorListInputs.meshPath, loadAccessorListInputs.buffers[accessor.mBufferIdx], pathJoint);
+    std::ostringstream name;
+    name << attributeName << setIndex++;
+    std::vector<uint8_t> buffer;
+    ReadTypedVectorAccessor<needsNormalize>(loadDataType, accessor, dataStream, buffer);
+    loadAccessorListInputs.rawData.mAttribs.push_back({name.str(), Property::VECTOR4, static_cast<uint32_t>(buffer.size() / sizeof(Vector4)), std::move(buffer)});
+  }
 }
 
-MeshDefinition::SparseBlob::SparseBlob(Blob&& indices, Blob&& values, uint32_t count)
-: mIndices(std::move(indices)),
-  mValues(std::move(values)),
-  mCount{count}
+void LoadIndices(LoadAccessorInputs indicesInput)
 {
-}
+  if(indicesInput.accessor.IsDefined())
+  {
+    if(MaskMatch(indicesInput.flags, MeshDefinition::Flags::U32_INDICES))
+    {
+      DALI_ASSERT_ALWAYS(((indicesInput.accessor.mBlob.mLength % sizeof(uint32_t) == 0) ||
+                          indicesInput.accessor.mBlob.mStride >= sizeof(uint32_t)) &&
+                         "Index buffer length not a multiple of element size");
+      const auto indexCount = indicesInput.accessor.mBlob.GetBufferSize() / sizeof(uint32_t);
+      indicesInput.rawData.mIndices.resize(indexCount * 2); // NOTE: we need space for uint32_ts initially.
 
-MeshDefinition::Accessor::Accessor(const MeshDefinition::Blob&       blob,
-                                   const MeshDefinition::SparseBlob& sparse,
-                                   Index                             bufferIndex)
-: mBlob{blob},
-  mSparse{(sparse.mIndices.IsDefined() && sparse.mValues.IsDefined()) ? new SparseBlob{sparse} : nullptr},
-  mBufferIdx(bufferIndex)
-{
-}
+      std::string path;
+      auto&       stream = GetAvailableData(indicesInput.fileStream, indicesInput.meshPath, indicesInput.buffers[indicesInput.accessor.mBufferIdx], path);
+      if(!ReadAccessor(indicesInput.accessor, stream, reinterpret_cast<uint8_t*>(indicesInput.rawData.mIndices.data())))
+      {
+        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indices from '" << path << "'.";
+      }
+    }
+    else if(MaskMatch(indicesInput.flags, MeshDefinition::Flags::U8_INDICES))
+    {
+      DALI_ASSERT_ALWAYS(((indicesInput.accessor.mBlob.mLength % sizeof(uint8_t) == 0) ||
+                          indicesInput.accessor.mBlob.mStride >= sizeof(uint8_t)) &&
+                         "Index buffer length not a multiple of element size");
+      const auto indexCount = indicesInput.accessor.mBlob.GetBufferSize() / sizeof(uint8_t);
+      indicesInput.rawData.mIndices.resize(indexCount); // NOTE: we need space for uint16_ts initially.
 
-MeshDefinition::Accessor::Accessor(MeshDefinition::Blob&&       blob,
-                                   MeshDefinition::SparseBlob&& sparse,
-                                   Index                        bufferIndex)
-: mBlob{std::move(blob)},
-  mSparse{(sparse.mIndices.IsDefined() && sparse.mValues.IsDefined()) ? new SparseBlob{std::move(sparse)} : nullptr},
-  mBufferIdx(bufferIndex)
-{
-}
+      std::string path;
+      auto        u8s    = reinterpret_cast<uint8_t*>(indicesInput.rawData.mIndices.data()) + indexCount;
+      auto&       stream = GetAvailableData(indicesInput.fileStream, indicesInput.meshPath, indicesInput.buffers[indicesInput.accessor.mBufferIdx], path);
+      if(!ReadAccessor(indicesInput.accessor, stream, u8s))
+      {
+        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indices from '" << path << "'.";
+      }
 
-void MeshDefinition::Blob::ComputeMinMax(std::vector<float>& min, std::vector<float>& max, uint32_t numComponents, uint32_t count, const float* values)
-{
-  min.assign(numComponents, MAXFLOAT);
-  max.assign(numComponents, -MAXFLOAT);
-  for(uint32_t i = 0; i < count; ++i)
-  {
-    for(uint32_t j = 0; j < numComponents; ++j)
+      auto u16s = indicesInput.rawData.mIndices.data();
+      auto end  = u8s + indexCount;
+      while(u8s != end)
+      {
+        *u16s = static_cast<uint16_t>(*u8s);
+        ++u16s;
+        ++u8s;
+      }
+    }
+    else
     {
-      min[j] = std::min(min[j], *values);
-      max[j] = std::max(max[j], *values);
-      values++;
+      DALI_ASSERT_ALWAYS(((indicesInput.accessor.mBlob.mLength % sizeof(unsigned short) == 0) ||
+                          indicesInput.accessor.mBlob.mStride >= sizeof(unsigned short)) &&
+                         "Index buffer length not a multiple of element size");
+      indicesInput.rawData.mIndices.resize(indicesInput.accessor.mBlob.mLength / sizeof(unsigned short));
+
+      std::string path;
+      auto&       stream = GetAvailableData(indicesInput.fileStream, indicesInput.meshPath, indicesInput.buffers[indicesInput.accessor.mBufferIdx], path);
+      if(!ReadAccessor(indicesInput.accessor, stream, reinterpret_cast<uint8_t*>(indicesInput.rawData.mIndices.data())))
+      {
+        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indicesInput.accessor from '" << path << "'.";
+      }
     }
   }
 }
 
-void MeshDefinition::Blob::ApplyMinMax(const std::vector<float>& min, const std::vector<float>& max, uint32_t count, float* values)
+uint32_t LoadPositions(LoadAccessorInputs positionsInput, bool hasBlendShape)
 {
-  DALI_ASSERT_DEBUG(max.size() == min.size() || max.size() * min.size() == 0);
-  const auto numComponents = std::max(min.size(), max.size());
-
-  using ClampFn   = void (*)(const float*, const float*, uint32_t, float&);
-  ClampFn clampFn = min.empty() ? (max.empty() ? static_cast<ClampFn>(nullptr) : [](const float* min, const float* max, uint32_t i, float& value)
-                                     { value = std::min(max[i], value); })
-                                : (max.empty() ? [](const float* min, const float* max, uint32_t i, float& value)
-                                     { value = std::max(min[i], value); }
-                                               : static_cast<ClampFn>([](const float* min, const float* max, uint32_t i, float& value)
-                                                                      { value = std::min(std::max(min[i], value), max[i]); }));
-
-  if(!clampFn)
-  {
-    return;
-  }
-
-  auto end = values + count * numComponents;
-  while(values != end)
-  {
-    auto     nextElement = values + numComponents;
-    uint32_t i           = 0;
-    while(values != nextElement)
-    {
-      clampFn(min.data(), max.data(), i, *values);
-      ++values;
-      ++i;
-    }
-  }
-}
-
-MeshDefinition::Blob::Blob(uint32_t offset, uint32_t length, uint16_t stride, uint16_t elementSizeHint, const std::vector<float>& min, const std::vector<float>& max)
-: mOffset(offset),
-  mLength(length),
-  mStride(stride),
-  mElementSizeHint(elementSizeHint),
-  mMin(min),
-  mMax(max)
-{
-}
-
-uint32_t MeshDefinition::Blob::GetBufferSize() const
-{
-  return mLength;
-}
-
-void MeshDefinition::Blob::ComputeMinMax(uint32_t numComponents, uint32_t count, float* values)
-{
-  ComputeMinMax(mMin, mMax, numComponents, count, values);
-}
-
-void MeshDefinition::Blob::ApplyMinMax(uint32_t count, float* values) const
-{
-  ApplyMinMax(mMin, mMax, count, values);
-}
-
-void MeshDefinition::RawData::Attrib::AttachBuffer(Geometry& g) const
-{
-  Property::Map attribMap;
-  attribMap[mName]          = mType;
-  VertexBuffer attribBuffer = VertexBuffer::New(attribMap);
-  attribBuffer.SetData(mData.data(), mNumElements);
-
-  g.AddVertexBuffer(attribBuffer);
-}
-
-bool MeshDefinition::IsQuad() const
-{
-  return CaseInsensitiveStringCompare(QUAD, mUri);
-}
-
-bool MeshDefinition::IsSkinned() const
-{
-  return mJoints0.IsDefined() && mWeights0.IsDefined();
-}
-
-bool MeshDefinition::HasBlendShapes() const
-{
-  return !mBlendShapes.empty();
-}
-
-void MeshDefinition::RequestNormals()
-{
-  mNormals.mBlob.mLength = mPositions.mBlob.GetBufferSize();
-}
-
-void MeshDefinition::RequestTangents()
-{
-  mTangents.mBlob.mLength = mNormals.mBlob.GetBufferSize();
-}
-
-MeshDefinition::RawData
-MeshDefinition::LoadRaw(const std::string& modelsPath, BufferDefinition::Vector& buffers)
-{
-  RawData raw;
-  if(IsQuad())
-  {
-    return raw;
-  }
-
-  std::string meshPath;
-  meshPath = modelsPath + mUri;
-  std::fstream fileStream;
-  if(!mUri.empty())
+  uint32_t             numVector3 = 0u;
+  std::vector<Vector3> positions;
+  if(positionsInput.accessor.IsDefined())
   {
-    fileStream.open(meshPath, std::ios::in | std::ios::binary);
-    if(!fileStream.is_open())
-    {
-      DALI_LOG_ERROR("Fail to open buffer from %s.\n", meshPath.c_str());
-    }
-  }
+    const auto bufferSize = positionsInput.accessor.mBlob.GetBufferSize();
 
-  if(mIndices.IsDefined())
-  {
-    if(MaskMatch(mFlags, U32_INDICES))
+    if(MaskMatch(positionsInput.flags, MeshDefinition::Flags::S8_POSITION) || MaskMatch(positionsInput.flags, MeshDefinition::Flags::U8_POSITION))
     {
-      DALI_ASSERT_ALWAYS(((mIndices.mBlob.mLength % sizeof(uint32_t) == 0) ||
-                          mIndices.mBlob.mStride >= sizeof(uint32_t)) &&
-                         "Index buffer length not a multiple of element size");
-      const auto indexCount = mIndices.mBlob.GetBufferSize() / sizeof(uint32_t);
-      raw.mIndices.resize(indexCount * 2); // NOTE: we need space for uint32_ts initially.
-
-      std::string path;
-      auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mIndices.mBufferIdx], path);
-      if(!ReadAccessor(mIndices, stream, reinterpret_cast<uint8_t*>(raw.mIndices.data())))
-      {
-        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indices from '" << path << "'.";
-      }
-
-      auto u16s = raw.mIndices.data();
-      auto u32s = reinterpret_cast<uint32_t*>(raw.mIndices.data());
-      auto end  = u32s + indexCount;
-      while(u32s != end)
-      {
-        *u16s = static_cast<uint16_t>(*u32s);
-        ++u16s;
-        ++u32s;
-      }
-
-      raw.mIndices.resize(indexCount);
+      DALI_ASSERT_ALWAYS(((positionsInput.accessor.mBlob.mLength % (sizeof(uint8_t) * 3) == 0) ||
+                          positionsInput.accessor.mBlob.mStride >= (sizeof(uint8_t) * 3)) &&
+                         "Position buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(uint8_t) * 3));
     }
-    else if(MaskMatch(mFlags, U8_INDICES))
+    else if(MaskMatch(positionsInput.flags, MeshDefinition::Flags::S16_POSITION) || MaskMatch(positionsInput.flags, MeshDefinition::Flags::U16_POSITION))
     {
-      DALI_ASSERT_ALWAYS(((mIndices.mBlob.mLength % sizeof(uint8_t) == 0) ||
-                          mIndices.mBlob.mStride >= sizeof(uint8_t)) &&
-                         "Index buffer length not a multiple of element size");
-      const auto indexCount = mIndices.mBlob.GetBufferSize() / sizeof(uint8_t);
-      raw.mIndices.resize(indexCount); // NOTE: we need space for uint32_ts initially.
-
-      std::string path;
-      auto u8s = reinterpret_cast<uint8_t*>(raw.mIndices.data()) + indexCount;
-      auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mIndices.mBufferIdx], path);
-      if(!ReadAccessor(mIndices, stream, u8s))
-      {
-        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indices from '" << path << "'.";
-      }
-
-      auto u16s = raw.mIndices.data();
-      auto end  = u8s + indexCount;
-      while(u8s != end)
-      {
-        *u16s = static_cast<uint16_t>(*u8s);
-        ++u16s;
-        ++u8s;
-      }
+      DALI_ASSERT_ALWAYS(((positionsInput.accessor.mBlob.mLength % (sizeof(uint16_t) * 3) == 0) ||
+                          positionsInput.accessor.mBlob.mStride >= (sizeof(uint16_t) * 3)) &&
+                         "Position buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(uint16_t) * 3));
     }
     else
     {
-      DALI_ASSERT_ALWAYS(((mIndices.mBlob.mLength % sizeof(unsigned short) == 0) ||
-                          mIndices.mBlob.mStride >= sizeof(unsigned short)) &&
-                         "Index buffer length not a multiple of element size");
-      raw.mIndices.resize(mIndices.mBlob.mLength / sizeof(unsigned short));
-
-
-      std::string path;
-      auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mIndices.mBufferIdx], path);
-      if(!ReadAccessor(mIndices, stream, reinterpret_cast<uint8_t*>(raw.mIndices.data())))
-      {
-        ExceptionFlinger(ASSERT_LOCATION) << "Failed to read indices from '" << path << "'.";
-      }
+      DALI_ASSERT_ALWAYS(((positionsInput.accessor.mBlob.mLength % sizeof(Vector3) == 0) ||
+                          positionsInput.accessor.mBlob.mStride >= sizeof(Vector3)) &&
+                         "Position buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
     }
-  }
 
-  std::vector<Vector3> positions;
-  if(mPositions.IsDefined())
-  {
-    DALI_ASSERT_ALWAYS(((mPositions.mBlob.mLength % sizeof(Vector3) == 0) ||
-                        mPositions.mBlob.mStride >= sizeof(Vector3)) &&
-                       "Position buffer length not a multiple of element size");
-    const auto           bufferSize = mPositions.mBlob.GetBufferSize();
     std::vector<uint8_t> buffer(bufferSize);
 
     std::string path;
-    auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mPositions.mBufferIdx], path);
-    if(!ReadAccessor(mPositions, stream, buffer.data()))
+    auto&       stream = GetAvailableData(positionsInput.fileStream, positionsInput.meshPath, positionsInput.buffers[positionsInput.accessor.mBufferIdx], path);
+    if(!ReadAccessor(positionsInput.accessor, stream, buffer.data()))
     {
       ExceptionFlinger(ASSERT_LOCATION) << "Failed to read positions from '" << path << "'.";
     }
 
-    uint32_t numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
-    if(mPositions.mBlob.mMin.size() != 3u || mPositions.mBlob.mMax.size() != 3u)
+    GetDequantizedData(buffer, 3u, numVector3, positionsInput.flags & MeshDefinition::FlagMasks::POSITIONS_MASK, positionsInput.accessor.mNormalized);
+
+    if(positionsInput.accessor.mNormalized)
     {
-      mPositions.mBlob.ComputeMinMax(3u, numVector3, reinterpret_cast<float*>(buffer.data()));
+      GetDequantizedMinMax(positionsInput.accessor.mBlob.mMin, positionsInput.accessor.mBlob.mMax, positionsInput.flags & MeshDefinition::FlagMasks::POSITIONS_MASK);
+    }
+
+    if(positionsInput.accessor.mBlob.mMin.size() != 3u || positionsInput.accessor.mBlob.mMax.size() != 3u)
+    {
+      positionsInput.accessor.mBlob.ComputeMinMax(3u, numVector3, reinterpret_cast<float*>(buffer.data()));
     }
     else
     {
-      mPositions.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()));
+      positionsInput.accessor.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()));
     }
 
-    if(HasBlendShapes())
+    if(hasBlendShape)
     {
       positions.resize(numVector3);
       std::copy(buffer.data(), buffer.data() + buffer.size(), reinterpret_cast<uint8_t*>(positions.data()));
     }
 
-    raw.mAttribs.push_back({"aPosition", Property::VECTOR3, numVector3, std::move(buffer)});
+    positionsInput.rawData.mAttribs.push_back({"aPosition", Property::VECTOR3, numVector3, std::move(buffer)});
   }
+  return numVector3;
+}
 
-  const auto isTriangles = mPrimitiveType == Geometry::TRIANGLES;
-  auto       hasNormals  = mNormals.IsDefined();
+bool LoadNormals(LoadAccessorInputs normalsInput, bool isTriangles, uint32_t positionBufferSize)
+{
+  auto hasNormals = normalsInput.accessor.IsDefined();
   if(hasNormals)
   {
-    DALI_ASSERT_ALWAYS(((mNormals.mBlob.mLength % sizeof(Vector3) == 0) ||
-                        mNormals.mBlob.mStride >= sizeof(Vector3)) &&
-                       "Normal buffer length not a multiple of element size");
-    const auto           bufferSize = mNormals.mBlob.GetBufferSize();
+    const auto bufferSize = normalsInput.accessor.mBlob.GetBufferSize();
+    uint32_t   numVector3;
+
+    if(MaskMatch(normalsInput.flags, MeshDefinition::Flags::S8_NORMAL))
+    {
+      DALI_ASSERT_ALWAYS(((normalsInput.accessor.mBlob.mLength % (sizeof(int8_t) * 3) == 0) ||
+                          normalsInput.accessor.mBlob.mStride >= (sizeof(int8_t) * 3)) &&
+                         "Normal buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int8_t) * 3));
+    }
+    else if(MaskMatch(normalsInput.flags, MeshDefinition::Flags::S16_NORMAL))
+    {
+      DALI_ASSERT_ALWAYS(((normalsInput.accessor.mBlob.mLength % (sizeof(int16_t) * 3) == 0) ||
+                          normalsInput.accessor.mBlob.mStride >= (sizeof(int16_t) * 3)) &&
+                         "Normal buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / (sizeof(int16_t) * 3));
+    }
+    else
+    {
+      DALI_ASSERT_ALWAYS(((normalsInput.accessor.mBlob.mLength % sizeof(Vector3) == 0) ||
+                          normalsInput.accessor.mBlob.mStride >= sizeof(Vector3)) &&
+                         "Normal buffer length not a multiple of element size");
+      numVector3 = static_cast<uint32_t>(bufferSize / sizeof(Vector3));
+    }
+
     std::vector<uint8_t> buffer(bufferSize);
 
     std::string path;
-    auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mNormals.mBufferIdx], path);
-    if(!ReadAccessor(mNormals, stream, buffer.data()))
+    auto&       stream = GetAvailableData(normalsInput.fileStream, normalsInput.meshPath, normalsInput.buffers[normalsInput.accessor.mBufferIdx], path);
+    if(!ReadAccessor(normalsInput.accessor, stream, buffer.data()))
     {
       ExceptionFlinger(ASSERT_LOCATION) << "Failed to read normals from '" << path << "'.";
     }
 
-    mNormals.mBlob.ApplyMinMax(static_cast<uint32_t>(bufferSize / sizeof(Vector3)), reinterpret_cast<float*>(buffer.data()));
+    GetDequantizedData(buffer, 3u, numVector3, normalsInput.flags & MeshDefinition::FlagMasks::NORMALS_MASK, normalsInput.accessor.mNormalized);
+
+    if(normalsInput.accessor.mNormalized)
+    {
+      GetDequantizedMinMax(normalsInput.accessor.mBlob.mMin, normalsInput.accessor.mBlob.mMax, normalsInput.flags & MeshDefinition::FlagMasks::NORMALS_MASK);
+    }
+
+    normalsInput.accessor.mBlob.ApplyMinMax(numVector3, reinterpret_cast<float*>(buffer.data()));
 
-    raw.mAttribs.push_back({"aNormal", Property::VECTOR3, static_cast<uint32_t>(bufferSize / sizeof(Vector3)), std::move(buffer)});
+    normalsInput.rawData.mAttribs.push_back({"aNormal", Property::VECTOR3, numVector3, std::move(buffer)});
   }
-  else if(mNormals.mBlob.mLength != 0 && isTriangles)
+  else if(normalsInput.accessor.mBlob.mLength != 0 && isTriangles)
   {
-    DALI_ASSERT_DEBUG(mNormals.mBlob.mLength == mPositions.mBlob.GetBufferSize());
-    GenerateNormals(raw);
-    hasNormals = true;
+    DALI_ASSERT_DEBUG(normalsInput.accessor.mBlob.mLength == positionBufferSize);
+    static const std::function<bool(MeshDefinition::RawData&)> GenerateNormalsFunction[2] =
+      {
+        GenerateNormals<false>,
+        GenerateNormals<true>,
+      };
+    const bool generateSuccessed = GenerateNormalsFunction[MaskMatch(normalsInput.flags, MeshDefinition::Flags::U32_INDICES)](normalsInput.rawData);
+    if(!generateSuccessed)
+    {
+      DALI_LOG_ERROR("Failed to generate normal\n");
+    }
+    else
+    {
+      hasNormals = true;
+    }
   }
+  return hasNormals;
+}
 
-  const auto hasUvs = mTexCoords.IsDefined();
-  if(hasUvs)
+void LoadTextureCoordinates(LoadAccessorListInputs textureCoordinatesInput)
+{
+  if(!textureCoordinatesInput.accessors.empty() && textureCoordinatesInput.accessors[0].IsDefined())
   {
-    DALI_ASSERT_ALWAYS(((mTexCoords.mBlob.mLength % sizeof(Vector2) == 0) ||
-                        mTexCoords.mBlob.mStride >= sizeof(Vector2)) &&
-                       "Normal buffer length not a multiple of element size");
-    const auto           bufferSize = mTexCoords.mBlob.GetBufferSize();
+    auto&      texCoords  = textureCoordinatesInput.accessors[0];
+    const auto bufferSize = texCoords.mBlob.GetBufferSize();
+    uint32_t   uvCount;
+
+    if(MaskMatch(textureCoordinatesInput.flags, MeshDefinition::Flags::S8_TEXCOORD) || MaskMatch(textureCoordinatesInput.flags, MeshDefinition::Flags::U8_TEXCOORD))
+    {
+      DALI_ASSERT_ALWAYS(((texCoords.mBlob.mLength % (sizeof(uint8_t) * 2) == 0) ||
+                          texCoords.mBlob.mStride >= (sizeof(uint8_t) * 2)) &&
+                         "TexCoords buffer length not a multiple of element size");
+      uvCount = static_cast<uint32_t>(bufferSize / (sizeof(uint8_t) * 2));
+    }
+    else if(MaskMatch(textureCoordinatesInput.flags, MeshDefinition::Flags::S16_TEXCOORD) || MaskMatch(textureCoordinatesInput.flags, MeshDefinition::Flags::U16_TEXCOORD))
+    {
+      DALI_ASSERT_ALWAYS(((texCoords.mBlob.mLength % (sizeof(uint16_t) * 2) == 0) ||
+                          texCoords.mBlob.mStride >= (sizeof(uint16_t) * 2)) &&
+                         "TexCoords buffer length not a multiple of element size");
+      uvCount = static_cast<uint32_t>(bufferSize / (sizeof(uint16_t) * 2));
+    }
+    else
+    {
+      DALI_ASSERT_ALWAYS(((texCoords.mBlob.mLength % sizeof(Vector2) == 0) ||
+                          texCoords.mBlob.mStride >= sizeof(Vector2)) &&
+                         "TexCoords buffer length not a multiple of element size");
+      uvCount = static_cast<uint32_t>(bufferSize / sizeof(Vector2));
+    }
+
     std::vector<uint8_t> buffer(bufferSize);
 
     std::string path;
-    auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mTexCoords.mBufferIdx], path);
-    if(!ReadAccessor(mTexCoords, stream, buffer.data()))
+    auto&       stream = GetAvailableData(textureCoordinatesInput.fileStream, textureCoordinatesInput.meshPath, textureCoordinatesInput.buffers[texCoords.mBufferIdx], path);
+    if(!ReadAccessor(texCoords, stream, buffer.data()))
     {
       ExceptionFlinger(ASSERT_LOCATION) << "Failed to read uv-s from '" << path << "'.";
     }
 
-    const auto uvCount = bufferSize / sizeof(Vector2);
-    if(MaskMatch(mFlags, FLIP_UVS_VERTICAL))
+    GetDequantizedData(buffer, 2u, uvCount, textureCoordinatesInput.flags & MeshDefinition::FlagMasks::TEXCOORDS_MASK, texCoords.mNormalized);
+
+    if(MaskMatch(textureCoordinatesInput.flags, MeshDefinition::Flags::FLIP_UVS_VERTICAL))
     {
       auto uv    = reinterpret_cast<Vector2*>(buffer.data());
       auto uvEnd = uv + uvCount;
@@ -810,148 +1089,208 @@ MeshDefinition::LoadRaw(const std::string& modelsPath, BufferDefinition::Vector&
       }
     }
 
-    mTexCoords.mBlob.ApplyMinMax(static_cast<uint32_t>(bufferSize / sizeof(Vector2)), reinterpret_cast<float*>(buffer.data()));
+    if(texCoords.mNormalized)
+    {
+      GetDequantizedMinMax(texCoords.mBlob.mMin, texCoords.mBlob.mMax, textureCoordinatesInput.flags & MeshDefinition::FlagMasks::TEXCOORDS_MASK);
+    }
 
-    raw.mAttribs.push_back({"aTexCoord", Property::VECTOR2, static_cast<uint32_t>(uvCount), std::move(buffer)});
+    texCoords.mBlob.ApplyMinMax(static_cast<uint32_t>(uvCount), reinterpret_cast<float*>(buffer.data()));
+    textureCoordinatesInput.rawData.mAttribs.push_back({"aTexCoord", Property::VECTOR2, static_cast<uint32_t>(uvCount), std::move(buffer)});
   }
+}
 
-  if(mTangents.IsDefined())
+void LoadTangents(LoadAccessorInputs tangentsInput, bool hasNormals, bool hasUvs, bool isTriangles, Property::Type tangentType, uint32_t normalBufferSize)
+{
+  if(tangentsInput.accessor.IsDefined())
   {
-    uint32_t propertySize = static_cast<uint32_t>((mTangentType == Property::VECTOR4) ? sizeof(Vector4) : sizeof(Vector3));
-    DALI_ASSERT_ALWAYS(((mTangents.mBlob.mLength % propertySize == 0) ||
-                        mTangents.mBlob.mStride >= propertySize) &&
-                       "Tangents buffer length not a multiple of element size");
-    const auto           bufferSize = mTangents.mBlob.GetBufferSize();
+    const auto bufferSize = tangentsInput.accessor.mBlob.GetBufferSize();
+
+    uint32_t propertySize   = static_cast<uint32_t>((tangentType == Property::VECTOR4) ? sizeof(Vector4) : sizeof(Vector3));
+    uint32_t componentCount = static_cast<uint32_t>(propertySize / sizeof(float));
+
+    uint32_t numTangents;
+
+    if(MaskMatch(tangentsInput.flags, MeshDefinition::Flags::S8_TANGENT))
+    {
+      DALI_ASSERT_ALWAYS(((tangentsInput.accessor.mBlob.mLength % (sizeof(int8_t) * componentCount) == 0) ||
+                          tangentsInput.accessor.mBlob.mStride >= (sizeof(int8_t) * componentCount)) &&
+                         "Tangents buffer length not a multiple of element size");
+      numTangents = static_cast<uint32_t>(bufferSize / (sizeof(int8_t) * componentCount));
+    }
+    else if(MaskMatch(tangentsInput.flags, MeshDefinition::Flags::S16_TANGENT))
+    {
+      DALI_ASSERT_ALWAYS(((tangentsInput.accessor.mBlob.mLength % (sizeof(int16_t) * componentCount) == 0) ||
+                          tangentsInput.accessor.mBlob.mStride >= (sizeof(int16_t) * componentCount)) &&
+                         "Tangents buffer length not a multiple of element size");
+      numTangents = static_cast<uint32_t>(bufferSize / (sizeof(int16_t) * componentCount));
+    }
+    else
+    {
+      DALI_ASSERT_ALWAYS(((tangentsInput.accessor.mBlob.mLength % propertySize == 0) ||
+                          tangentsInput.accessor.mBlob.mStride >= propertySize) &&
+                         "Tangents buffer length not a multiple of element size");
+      numTangents = static_cast<uint32_t>(bufferSize / propertySize);
+    }
+
     std::vector<uint8_t> buffer(bufferSize);
 
     std::string path;
-    auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mTangents.mBufferIdx], path);
-    if(!ReadAccessor(mTangents, stream, buffer.data()))
+    auto&       stream = GetAvailableData(tangentsInput.fileStream, tangentsInput.meshPath, tangentsInput.buffers[tangentsInput.accessor.mBufferIdx], path);
+    if(!ReadAccessor(tangentsInput.accessor, stream, buffer.data()))
     {
       ExceptionFlinger(ASSERT_LOCATION) << "Failed to read tangents from '" << path << "'.";
     }
-    mTangents.mBlob.ApplyMinMax(bufferSize / propertySize, reinterpret_cast<float*>(buffer.data()));
 
-    raw.mAttribs.push_back({"aTangent", mTangentType, static_cast<uint32_t>(bufferSize / propertySize), std::move(buffer)});
+    GetDequantizedData(buffer, componentCount, numTangents, tangentsInput.flags & MeshDefinition::FlagMasks::TANGENTS_MASK, tangentsInput.accessor.mNormalized);
+
+    if(tangentsInput.accessor.mNormalized)
+    {
+      GetDequantizedMinMax(tangentsInput.accessor.mBlob.mMin, tangentsInput.accessor.mBlob.mMax, tangentsInput.flags & MeshDefinition::FlagMasks::TANGENTS_MASK);
+    }
+
+    tangentsInput.accessor.mBlob.ApplyMinMax(numTangents, reinterpret_cast<float*>(buffer.data()));
+
+    tangentsInput.rawData.mAttribs.push_back({"aTangent", tangentType, static_cast<uint32_t>(numTangents), std::move(buffer)});
   }
-  else if(mTangents.mBlob.mLength != 0 && hasNormals && isTriangles)
+  else if(tangentsInput.accessor.mBlob.mLength != 0 && hasNormals && isTriangles)
   {
-    DALI_ASSERT_DEBUG(mTangents.mBlob.mLength == mNormals.mBlob.GetBufferSize());
-    hasUvs ? GenerateTangentsWithUvs(raw) : GenerateTangents(raw);
+    DALI_ASSERT_DEBUG(tangentsInput.accessor.mBlob.mLength == normalBufferSize);
+    static const std::function<bool(MeshDefinition::RawData&)> GenerateTangentsFunction[2][2][2] =
+      {
+        {
+          {
+            GenerateTangents<false, false, false>,
+            GenerateTangents<false, false, true>,
+          },
+          {
+            GenerateTangents<false, true, false>,
+            GenerateTangents<false, true, true>,
+          },
+        },
+        {
+          {
+            GenerateTangents<true, false, false>,
+            GenerateTangents<true, false, true>,
+          },
+          {
+            GenerateTangents<true, true, false>,
+            GenerateTangents<true, true, true>,
+          },
+        }};
+    const bool generateSuccessed = GenerateTangentsFunction[MaskMatch(tangentsInput.flags, MeshDefinition::Flags::U32_INDICES)][tangentType == Property::VECTOR3][hasUvs](tangentsInput.rawData);
+    if(!generateSuccessed)
+    {
+      DALI_LOG_ERROR("Failed to generate tangents\n");
+    }
   }
+}
 
-  if(mColors.IsDefined())
+void LoadColors(LoadAccessorListInputs colorsInput)
+{
+  // Only support 1 vertex color
+  if(!colorsInput.accessors.empty() && colorsInput.accessors[0].IsDefined())
   {
-    uint32_t       propertySize = mColors.mBlob.mElementSizeHint;
+    uint32_t       propertySize = colorsInput.accessors[0].mBlob.mElementSizeHint;
     Property::Type propertyType = (propertySize == sizeof(Vector4)) ? Property::VECTOR4 : ((propertySize == sizeof(Vector3)) ? Property::VECTOR3 : Property::NONE);
     if(propertyType != Property::NONE)
     {
-      DALI_ASSERT_ALWAYS(((mColors.mBlob.mLength % propertySize == 0) ||
-                          mColors.mBlob.mStride >= propertySize) &&
+      DALI_ASSERT_ALWAYS(((colorsInput.accessors[0].mBlob.mLength % propertySize == 0) ||
+                          colorsInput.accessors[0].mBlob.mStride >= propertySize) &&
                          "Colors buffer length not a multiple of element size");
-      const auto           bufferSize = mColors.mBlob.GetBufferSize();
+      const auto           bufferSize = colorsInput.accessors[0].mBlob.GetBufferSize();
       std::vector<uint8_t> buffer(bufferSize);
 
       std::string path;
-      auto&       stream = GetAvailableData(fileStream, meshPath, buffers[mColors.mBufferIdx], path);
-      if(!ReadAccessor(mColors, stream, buffer.data()))
+      auto&       stream = GetAvailableData(colorsInput.fileStream, colorsInput.meshPath, colorsInput.buffers[colorsInput.accessors[0].mBufferIdx], path);
+      if(!ReadAccessor(colorsInput.accessors[0], stream, buffer.data()))
       {
         ExceptionFlinger(ASSERT_LOCATION) << "Failed to read colors from '" << path << "'.";
       }
-      mColors.mBlob.ApplyMinMax(bufferSize / propertySize, reinterpret_cast<float*>(buffer.data()));
+      colorsInput.accessors[0].mBlob.ApplyMinMax(bufferSize / propertySize, reinterpret_cast<float*>(buffer.data()));
 
-      raw.mAttribs.push_back({"aVertexColor", propertyType, static_cast<uint32_t>(bufferSize / propertySize), std::move(buffer)});
+      colorsInput.rawData.mAttribs.push_back({"aVertexColor", propertyType, static_cast<uint32_t>(bufferSize / propertySize), std::move(buffer)});
     }
   }
-
-  if(IsSkinned())
+  else
   {
-    std::string pathJoint;
-    auto&       streamJoint = GetAvailableData(fileStream, meshPath, buffers[mJoints0.mBufferIdx], pathJoint);
-    if(MaskMatch(mFlags, U16_JOINT_IDS))
-    {
-      ReadJointAccessor<uint16_t>(raw, mJoints0, streamJoint, pathJoint);
-    }
-    else if(MaskMatch(mFlags, U8_JOINT_IDS))
-    {
-      ReadJointAccessor<uint8_t>(raw, mJoints0, streamJoint, pathJoint);
-    }
-    else
-    {
-      ReadJointAccessor<float>(raw, mJoints0, streamJoint, pathJoint);
-    }
+    std::vector<uint8_t> buffer(colorsInput.rawData.mAttribs[0].mNumElements * sizeof(Vector4));
+    auto                 colors = reinterpret_cast<Vector4*>(buffer.data());
 
-    DALI_ASSERT_ALWAYS(((mWeights0.mBlob.mLength % sizeof(Vector4) == 0) ||
-                        mWeights0.mBlob.mStride >= sizeof(Vector4)) &&
-                       "Weights buffer length not a multiple of element size");
-    const auto           bufferSize = mWeights0.mBlob.GetBufferSize();
-    std::vector<uint8_t> buffer(bufferSize);
-
-    std::string pathWeight;
-    auto&       streamWeight = GetAvailableData(fileStream, meshPath, buffers[mWeights0.mBufferIdx], pathWeight);
-    if(!ReadAccessor(mWeights0, streamWeight, buffer.data()))
+    for(uint32_t i = 0; i < colorsInput.rawData.mAttribs[0].mNumElements; i++)
     {
-      ExceptionFlinger(ASSERT_LOCATION) << "Failed to read weights from '" << pathWeight << "'.";
+      colors[i] = Vector4::ONE;
     }
 
-    raw.mAttribs.push_back({"aWeights", Property::VECTOR4, static_cast<uint32_t>(bufferSize / sizeof(Vector4)), std::move(buffer)});
+    colorsInput.rawData.mAttribs.push_back({"aVertexColor", Property::VECTOR4, colorsInput.rawData.mAttribs[0].mNumElements, std::move(buffer)});
   }
+}
 
+void LoadBlendShapes(MeshDefinition::RawData& rawData, std::vector<MeshDefinition::BlendShape>& blendShapes, MeshDefinition::Blob& blendShapeHeader, BlendShapes::Version blendShapeVersion, uint32_t numberOfVertices, std::fstream& fileStream, BufferDefinition::Vector& buffers)
+{
   // Calculate the Blob for the blend shapes.
-  Blob blendShapesBlob;
+  MeshDefinition::Blob blendShapesBlob;
   blendShapesBlob.mOffset = std::numeric_limits<unsigned int>::max();
   blendShapesBlob.mLength = 0u;
 
-  for(const auto& blendShape : mBlendShapes)
-  {
-    for(auto i : {&blendShape.deltas, &blendShape.normals, &blendShape.tangents})
+  uint32_t totalTextureSize(0u);
+
+  auto processAccessor = [&](const MeshDefinition::Accessor& accessor, uint32_t vector3Size) {
+    if(accessor.IsDefined())
     {
-      if(i->IsDefined())
-      {
-        blendShapesBlob.mOffset = std::min(blendShapesBlob.mOffset, i->mBlob.mOffset);
-        blendShapesBlob.mLength += i->mBlob.mLength;
-      }
+      blendShapesBlob.mOffset = std::min(blendShapesBlob.mOffset, accessor.mBlob.mOffset);
+      blendShapesBlob.mLength += accessor.mBlob.mLength;
+
+      totalTextureSize += accessor.mBlob.mLength / vector3Size;
     }
-  }
+  };
 
-  if(HasBlendShapes())
+  for(const auto& blendShape : blendShapes)
   {
-    const uint32_t numberOfVertices = static_cast<uint32_t>(mPositions.mBlob.mLength / sizeof(Vector3));
+    const auto positionMask = blendShape.mFlags & MeshDefinition::FlagMasks::POSITIONS_MASK;
+    const auto normalMask   = blendShape.mFlags & MeshDefinition::FlagMasks::NORMALS_MASK;
+    const auto tangentMask  = blendShape.mFlags & MeshDefinition::FlagMasks::TANGENTS_MASK;
+
+    processAccessor(blendShape.deltas, MaskMatch(positionMask, MeshDefinition::S8_POSITION) ? sizeof(uint8_t) * 3 : (MaskMatch(positionMask, MeshDefinition::S16_POSITION) ? sizeof(uint16_t) * 3 : sizeof(Vector3)));
+    processAccessor(blendShape.normals, MaskMatch(normalMask, MeshDefinition::S8_NORMAL) ? sizeof(uint8_t) * 3 : (MaskMatch(normalMask, MeshDefinition::S16_NORMAL) ? sizeof(uint16_t) * 3 : sizeof(Vector3)));
+    processAccessor(blendShape.tangents, MaskMatch(tangentMask, MeshDefinition::S8_TANGENT) ? sizeof(uint8_t) * 3 : (MaskMatch(tangentMask, MeshDefinition::S16_TANGENT) ? sizeof(uint16_t) * 3 : sizeof(Vector3)));
+  }
 
+  if(!blendShapes.empty())
+  {
     // Calculate the size of one buffer inside the texture.
-    raw.mBlendShapeBufferOffset = numberOfVertices;
+    rawData.mBlendShapeBufferOffset = numberOfVertices;
 
     bool     calculateGltf2BlendShapes = false;
     uint32_t textureWidth              = 0u;
     uint32_t textureHeight             = 0u;
 
-    if(!mBlendShapeHeader.IsDefined())
+    if(!blendShapeHeader.IsDefined())
     {
-      CalculateTextureSize(static_cast<uint32_t>(blendShapesBlob.mLength / sizeof(Vector3)), textureWidth, textureHeight);
+      CalculateTextureSize(totalTextureSize, textureWidth, textureHeight);
       calculateGltf2BlendShapes = true;
     }
     else
     {
       uint16_t header[2u];
-      ReadBlob(mBlendShapeHeader, fileStream, reinterpret_cast<uint8_t*>(header));
+      ReadBlob(blendShapeHeader, fileStream, reinterpret_cast<uint8_t*>(header));
       textureWidth  = header[0u];
       textureHeight = header[1u];
     }
 
-    const uint32_t numberOfBlendShapes = mBlendShapes.size();
-    raw.mBlendShapeUnnormalizeFactor.Resize(numberOfBlendShapes);
+    const uint32_t numberOfBlendShapes = blendShapes.size();
+    rawData.mBlendShapeUnnormalizeFactor.Resize(numberOfBlendShapes);
 
     Devel::PixelBuffer geometryPixelBuffer = Devel::PixelBuffer::New(textureWidth, textureHeight, Pixel::RGB32F);
     uint8_t*           geometryBuffer      = geometryPixelBuffer.GetBuffer();
 
     if(calculateGltf2BlendShapes)
     {
-      CalculateGltf2BlendShapes(geometryBuffer, mBlendShapes, numberOfVertices, raw.mBlendShapeUnnormalizeFactor[0u], buffers);
+      CalculateGltf2BlendShapes(geometryBuffer, blendShapes, numberOfVertices, rawData.mBlendShapeUnnormalizeFactor[0u], buffers);
     }
     else
     {
-      Blob unnormalizeFactorBlob;
-      unnormalizeFactorBlob.mLength = static_cast<uint32_t>(sizeof(float) * ((BlendShapes::Version::VERSION_2_0 == mBlendShapeVersion) ? 1u : numberOfBlendShapes));
+      MeshDefinition::Blob unnormalizeFactorBlob;
+      unnormalizeFactorBlob.mLength = static_cast<uint32_t>(sizeof(float) * ((BlendShapes::Version::VERSION_2_0 == blendShapeVersion) ? 1u : numberOfBlendShapes));
 
       if(blendShapesBlob.IsDefined())
       {
@@ -964,12 +1303,225 @@ MeshDefinition::LoadRaw(const std::string& modelsPath, BufferDefinition::Vector&
       // Read the unnormalize factors.
       if(unnormalizeFactorBlob.IsDefined())
       {
-        ReadBlob(unnormalizeFactorBlob, fileStream, reinterpret_cast<uint8_t*>(&raw.mBlendShapeUnnormalizeFactor[0u]));
+        ReadBlob(unnormalizeFactorBlob, fileStream, reinterpret_cast<uint8_t*>(&rawData.mBlendShapeUnnormalizeFactor[0u]));
       }
     }
-    raw.mBlendShapeData = Devel::PixelBuffer::Convert(geometryPixelBuffer);
+    rawData.mBlendShapeData = Devel::PixelBuffer::Convert(geometryPixelBuffer);
+  }
+}
+
+constexpr uint32_t MINIMUM_SHADER_VERSION_SUPPORT_VERTEX_ID = 300;
+
+} // namespace
+
+MeshDefinition::SparseBlob::SparseBlob(const Blob& indices, const Blob& values, uint32_t count)
+: mIndices{indices},
+  mValues{values},
+  mCount{count}
+{
+}
+
+MeshDefinition::SparseBlob::SparseBlob(Blob&& indices, Blob&& values, uint32_t count)
+: mIndices(std::move(indices)),
+  mValues(std::move(values)),
+  mCount{count}
+{
+}
+
+MeshDefinition::Accessor::Accessor(const MeshDefinition::Blob&       blob,
+                                   const MeshDefinition::SparseBlob& sparse,
+                                   Index                             bufferIndex,
+                                   bool                              normalized)
+: mBlob{blob},
+  mSparse{(sparse.mIndices.IsDefined() && sparse.mValues.IsDefined()) ? new SparseBlob{sparse} : nullptr},
+  mBufferIdx(bufferIndex),
+  mNormalized(normalized)
+{
+}
+
+MeshDefinition::Accessor::Accessor(MeshDefinition::Blob&&       blob,
+                                   MeshDefinition::SparseBlob&& sparse,
+                                   Index                        bufferIndex,
+                                   bool                         normalized)
+: mBlob{std::move(blob)},
+  mSparse{(sparse.mIndices.IsDefined() && sparse.mValues.IsDefined()) ? new SparseBlob{std::move(sparse)} : nullptr},
+  mBufferIdx(bufferIndex),
+  mNormalized(normalized)
+{
+}
+
+void MeshDefinition::Blob::ComputeMinMax(std::vector<float>& min, std::vector<float>& max, uint32_t numComponents, uint32_t count, const float* values)
+{
+  min.assign(numComponents, MAXFLOAT);
+  max.assign(numComponents, -MAXFLOAT);
+  for(uint32_t i = 0; i < count; ++i)
+  {
+    for(uint32_t j = 0; j < numComponents; ++j)
+    {
+      min[j] = std::min(min[j], *values);
+      max[j] = std::max(max[j], *values);
+      values++;
+    }
+  }
+}
+
+void MeshDefinition::Blob::ApplyMinMax(const std::vector<float>& min, const std::vector<float>& max, uint32_t count, float* values, std::vector<uint32_t>* sparseIndices)
+{
+  DALI_ASSERT_DEBUG(max.size() == min.size() || max.size() * min.size() == 0);
+  const auto numComponents = std::max(min.size(), max.size());
+
+  using ClampFn   = void (*)(const float*, const float*, uint32_t, float&);
+  ClampFn clampFn = min.empty() ? (max.empty() ? static_cast<ClampFn>(nullptr) : [](const float* min, const float* max, uint32_t i, float& value) { value = std::min(max[i], value); })
+                                : (max.empty() ? [](const float* min, const float* max, uint32_t i, float& value) { value = std::max(min[i], value); }
+                                               : static_cast<ClampFn>([](const float* min, const float* max, uint32_t i, float& value) { value = std::min(std::max(min[i], value), max[i]); }));
+
+  if(!clampFn)
+  {
+    return;
+  }
+
+  auto end = values + count * numComponents;
+  while(values != end)
+  {
+    auto     nextElement = values + numComponents;
+    uint32_t i           = 0;
+    while(values != nextElement)
+    {
+      clampFn(min.data(), max.data(), i, *values);
+      ++values;
+      ++i;
+    }
+  }
+}
+
+MeshDefinition::Blob::Blob(uint32_t offset, uint32_t length, uint16_t stride, uint16_t elementSizeHint, const std::vector<float>& min, const std::vector<float>& max)
+: mOffset(offset),
+  mLength(length),
+  mStride(stride),
+  mElementSizeHint(elementSizeHint),
+  mMin(min),
+  mMax(max)
+{
+}
+
+uint32_t MeshDefinition::Blob::GetBufferSize() const
+{
+  return mLength;
+}
+
+void MeshDefinition::Blob::ComputeMinMax(uint32_t numComponents, uint32_t count, float* values)
+{
+  ComputeMinMax(mMin, mMax, numComponents, count, values);
+}
+
+void MeshDefinition::Blob::ApplyMinMax(uint32_t count, float* values, std::vector<uint32_t>* sparseIndices) const
+{
+  ApplyMinMax(mMin, mMax, count, values, sparseIndices);
+}
+
+void MeshDefinition::RawData::Attrib::AttachBuffer(Geometry& g) const
+{
+  Property::Map attribMap;
+  attribMap[mName]          = mType;
+  VertexBuffer attribBuffer = VertexBuffer::New(attribMap);
+  attribBuffer.SetData(mData.data(), mNumElements);
+
+  g.AddVertexBuffer(attribBuffer);
+}
+
+bool MeshDefinition::IsQuad() const
+{
+  return CaseInsensitiveStringCompare(QUAD, mUri);
+}
+
+bool MeshDefinition::IsSkinned() const
+{
+  return !mJoints.empty() && !mWeights.empty();
+}
+
+bool MeshDefinition::HasVertexColor() const
+{
+  return !mColors.empty();
+}
+
+uint32_t MeshDefinition::GetNumberOfJointSets() const
+{
+  uint32_t number = static_cast<uint32_t>(mJoints.size());
+  if(number > MeshDefinition::MAX_NUMBER_OF_JOINT_SETS)
+  {
+    number = MeshDefinition::MAX_NUMBER_OF_JOINT_SETS;
+  }
+  return number;
+}
+
+bool MeshDefinition::HasBlendShapes() const
+{
+  return !mBlendShapes.empty();
+}
+
+void MeshDefinition::RequestNormals()
+{
+  mNormals.mBlob.mLength = mPositions.mBlob.GetBufferSize();
+}
+
+void MeshDefinition::RequestTangents()
+{
+  mTangents.mBlob.mLength = mNormals.mBlob.GetBufferSize();
+}
+
+MeshDefinition::RawData
+MeshDefinition::LoadRaw(const std::string& modelsPath, BufferDefinition::Vector& buffers)
+{
+  RawData raw;
+  if(IsQuad())
+  {
+    return raw;
+  }
+
+  std::string meshPath;
+  meshPath = modelsPath + mUri;
+  std::fstream fileStream;
+  if(!mUri.empty())
+  {
+    fileStream.open(meshPath, std::ios::in | std::ios::binary);
+    if(!fileStream.is_open())
+    {
+      DALI_LOG_ERROR("Fail to open buffer from %s.\n", meshPath.c_str());
+    }
+  }
+
+  LoadAccessorInputs indicesInput = {raw, mIndices, mFlags, fileStream, meshPath, buffers};
+  LoadIndices(indicesInput);
+
+  LoadAccessorInputs positionsInput   = {raw, mPositions, mFlags, fileStream, meshPath, buffers};
+  uint32_t           numberOfVertices = LoadPositions(positionsInput, HasBlendShapes());
+
+  const auto         isTriangles  = mPrimitiveType == Geometry::TRIANGLES;
+  LoadAccessorInputs normalsInput = {raw, mNormals, mFlags, fileStream, meshPath, buffers};
+  auto               hasNormals   = LoadNormals(normalsInput, isTriangles, mPositions.mBlob.GetBufferSize());
+
+  LoadAccessorListInputs textureCoordinatesInput = {raw, mTexCoords, mFlags, fileStream, meshPath, buffers};
+  LoadTextureCoordinates(textureCoordinatesInput);
+
+  const bool         hasUvs        = !mTexCoords.empty() && mTexCoords[0].IsDefined();
+  LoadAccessorInputs tangentsInput = {raw, mTangents, mFlags, fileStream, meshPath, buffers};
+  LoadTangents(tangentsInput, hasNormals, hasUvs, isTriangles, mTangentType, mNormals.mBlob.GetBufferSize());
+
+  LoadAccessorListInputs colorsInput = {raw, mColors, mFlags, fileStream, meshPath, buffers};
+  LoadColors(colorsInput);
+
+  if(IsSkinned())
+  {
+    LoadDataType           loadDataType = (MaskMatch(mFlags, MeshDefinition::U16_JOINT_IDS)) ? LoadDataType::UNSIGNED_SHORT : (MaskMatch(mFlags, MeshDefinition::U8_JOINT_IDS) ? LoadDataType::UNSIGNED_BYTE : LoadDataType::FLOAT);
+    LoadAccessorListInputs jointsInput  = {raw, mJoints, mFlags, fileStream, meshPath, buffers};
+    ReadTypedVectorAccessors<false>(jointsInput, loadDataType, "aJoints");
+
+    loadDataType                        = (MaskMatch(mFlags, MeshDefinition::U16_WEIGHT)) ? LoadDataType::UNSIGNED_SHORT : (MaskMatch(mFlags, MeshDefinition::U8_WEIGHT) ? LoadDataType::UNSIGNED_BYTE : LoadDataType::FLOAT);
+    LoadAccessorListInputs weightsInput = {raw, mWeights, mFlags, fileStream, meshPath, buffers};
+    ReadTypedVectorAccessors<true>(weightsInput, loadDataType, "aWeights");
   }
 
+  LoadBlendShapes(raw, mBlendShapes, mBlendShapeHeader, mBlendShapeVersion, numberOfVertices, fileStream, buffers);
   return raw;
 }
 
@@ -988,7 +1540,38 @@ MeshGeometry MeshDefinition::Load(RawData&& raw) const
   {
     if(!raw.mIndices.empty())
     {
-      meshGeometry.geometry.SetIndexBuffer(raw.mIndices.data(), raw.mIndices.size());
+      if(MaskMatch(mFlags, U32_INDICES))
+      {
+        // TODO : We can only store indeces as uint16_type. Send Dali::Geometry that we use it as uint32_t actual.
+        meshGeometry.geometry.SetIndexBuffer(reinterpret_cast<const uint32_t*>(raw.mIndices.data()), raw.mIndices.size() / 2);
+      }
+      else
+      {
+        meshGeometry.geometry.SetIndexBuffer(raw.mIndices.data(), raw.mIndices.size());
+      }
+    }
+
+    if(DALI_UNLIKELY(Dali::Shader::GetShaderLanguageVersion() < MINIMUM_SHADER_VERSION_SUPPORT_VERTEX_ID && !raw.mAttribs.empty()))
+    {
+      auto numElements = raw.mAttribs[0].mNumElements;
+
+      // gl_VertexID not support. We should add buffer hard.
+      Property::Map attribMap;
+      attribMap["aVertexID"] = Property::FLOAT;
+
+      VertexBuffer attribBuffer = VertexBuffer::New(attribMap);
+
+      std::vector<uint8_t> buffer(numElements * sizeof(float));
+      auto                 ids = reinterpret_cast<float*>(buffer.data());
+
+      for(uint32_t i = 0; i < numElements; i++)
+      {
+        ids[i] = static_cast<float>(i);
+      }
+
+      attribBuffer.SetData(buffer.data(), numElements);
+
+      meshGeometry.geometry.AddVertexBuffer(attribBuffer);
     }
 
     for(auto& a : raw.mAttribs)
@@ -1012,6 +1595,14 @@ MeshGeometry MeshDefinition::Load(RawData&& raw) const
   return meshGeometry;
 }
 
-} // namespace Loader
-} // namespace Scene3D
-} // namespace Dali
+void MeshDefinition::RetrieveBlendShapeComponents(bool& hasPositions, bool& hasNormals, bool& hasTangents) const
+{
+  for(const auto& blendShape : mBlendShapes)
+  {
+    hasPositions = hasPositions || blendShape.deltas.IsDefined();
+    hasNormals   = hasNormals || blendShape.normals.IsDefined();
+    hasTangents  = hasTangents || blendShape.tangents.IsDefined();
+  }
+}
+
+} // namespace Dali::Scene3D::Loader