src/armnnTfLiteParser/TensorFlowLiteSupport.md

   1 # TensorFlow Lite operators that the Arm NN SDK supports
   2
   3 This reference guide provides a list of TensorFlow Lite operators the Arm NN SDK currently supports.
   4
   5 ## Fully supported
   6
   7 The Arm NN SDK TensorFlow Lite parser currently supports the following operators:
   8
   9 * ADD
  10
  11 * AVERAGE_POOL_2D, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  12
  13 * BATCH_TO_SPACE
  14
  15 * CONCATENATION, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  16
  17 * CONV_2D, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  18
  19 * DEPTHWISE_CONV_2D, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  20
  21 * DEQUANTIZE
  22
  23 * EXP
  24
  25 * FULLY_CONNECTED, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  26
  27 * LOGISTIC
  28
  29 * L2_NORMALIZATION
  30
  31 * MAX_POOL_2D, Supported Fused Activation: RELU , RELU6 , TANH, NONE
  32
  33 * MAXIMUM
  34
  35 * MEAN
  36
  37 * MINIMUM
  38
  39 * MUL
  40
  41 * PACK
  42
  43 * PAD
  44
  45 * QUANTIZE
  46
  47 * RELU
  48
  49 * RELU6
  50
  51 * RESHAPE
  52
  53 * RESIZE_BILINEAR
  54
  55 * RESIZE_NEAREST_NEIGHBOR
  56
  57 * SLICE
  58
  59 * SOFTMAX
  60
  61 * SPACE_TO_BATCH
  62
  63 * SPLIT
  64
  65 * SPLIT_V
  66
  67 * SQUEEZE
  68
  69 * STRIDED_SLICE
  70
  71 * SUB
  72
  73 * TANH
  74
  75 * TRANSPOSE
  76
  77 * TRANSPOSE_CONV
  78
  79 * UNPACK
  80
  81 ## Custom Operator
  82
  83 * TFLite_Detection_PostProcess
  84
  85 ## Tested networks
  86
  87 Arm tested these operators with the following TensorFlow Lite neural network:
  88
  89 * [Quantized MobileNet](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224_quant.tgz)
  90
  91 * [Quantized SSD MobileNet](http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_quantized_300x300_coco14_sync_2018_07_18.tar.gz)
  92
  93 * DeepSpeech v1 converted from [TensorFlow model](https://github.com/mozilla/DeepSpeech/releases/tag/v0.4.1)
  94
  95 * DeepSpeaker
  96
  97 * [DeepLab v3+](https://www.tensorflow.org/lite/models/segmentation/overview)
  98
  99 * FSRCNN
 100
 101 * RDN converted from [TensorFlow model](https://github.com/hengchuan/RDN-TensorFlow)
 102
 103 * Quantized RDN (CpuRef)
 104
 105 * [Quantized Inception v3](http://download.tensorflow.org/models/tflite_11_05_08/inception_v3_quant.tgz)
 106
 107 * [Quantized Inception v4](http://download.tensorflow.org/models/inception_v4_299_quant_20181026.tgz) (CpuRef)
 108
 109 * Quantized ResNet v2 50 (CpuRef)
 110
 111 * Quantized Yolo v3 (CpuRef)
 112
 113 More machine learning operators will be supported in future releases.