3 [![Code Coverage](http://nnstreamer.mooo.com/nntrainer/ci/badge/codecoverage.svg)](http://nnstreamer.mooo.com/nntrainer/ci/gcov_html/index.html)
4 ![GitHub repo size](https://img.shields.io/github/repo-size/nnstreamer/nntrainer)
5 ![GitHub issues](https://img.shields.io/github/issues/nnstreamer/nntrainer)
6 ![GitHub pull requests](https://img.shields.io/github/issues-pr/nnstreamer/nntrainer)
7 <a href="https://scan.coverity.com/projects/nnstreamer-nntrainer">
8 <img alt="Coverity Scan Build Status"
9 src="https://scan.coverity.com/projects/22512/badge.svg"/>
11 [![DailyBuild](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/daily_build_test_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/)
13 NNtrainer is a Software Framework for training Neural Network models on devices.
17 NNtrainer is an Open Source Project. The aim of the NNtrainer is to develop a Software Framework to train neural network models on embedded devices which have relatively limited resources. Rather than training whole layers of a network, NNtrainer trains only one or a few layers of the layers added after a feature extractor.
19 Even though NNTrainer can be used to train sub-models, it requires implementation of additional functionalities to train models obtained from other machine learning and deep learning libraries. In the current version, various machine learning algorithms such as k-Nearest Neighbor (k-NN), Neural Networks, Logistic Regression and Reinforcement Learning algorithms are implemented. We also provide examples for various tasks such as transfer learning of models. In some of these examples, deep learning models such as Mobilenet V2 trained with Tensorflow-lite, are used as feature extractors. All of these were tested on Galaxy S8 with Android and PC (Ubuntu 16.04).
23 | | [Tizen](http://download.tizen.org/snapshots/tizen/unified/latest/repos/standard/packages/) | [Ubuntu](https://launchpad.net/~nnstreamer/+archive/ubuntu/ppa) | Android/NDK Build |
24 | :-- | :--: | :--: | :--: |
25 | | 6.0M2 and later | 18.04 | 9/P |
26 | arm | [![armv7l badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/tizen.armv7l_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) | Available | Ready |
27 | arm64 | [![aarch64 badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/tizen.aarch64_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) | Available | [![android badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/arm64_v8a_android_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) |
28 | x64 | [![x64 badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/tizen.x86_64_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) | [![ubuntu badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/ubuntu_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) | Ready |
29 | x86 | [![x86 badge](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/badge/tizen.i586_result_badge.svg)](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/) | N/A | N/A |
30 | Publish | [Tizen Repo](http://download.tizen.org/snapshots/tizen/unified/latest/repos/standard/packages/) | [PPA](https://launchpad.net/~nnstreamer/+archive/ubuntu/ppa) | |
31 | API | C (Official) | C/C++ | C/C++ |
33 - Ready: CI system ensures build-ability and unit-testing. Users may easily build and execute. However, we do not have automated release & deployment system for this instance.
34 - Available: binary packages are released and deployed automatically and periodically along with CI tests.
35 - [Daily Release](http://nnstreamer.mooo.com/nntrainer/ci/daily-build/build_result/)
36 - SDK Support: Tizen Studio (6.0 M2+)
39 * [Jijoong Moon](https://github.com/jijoongmoon)
40 * [MyungJoo Ham](https://github.com/myungjoo)
41 * [Geunsik Lim](https://github.com/leemgs)
44 * [Sangjung Woo](https://github.com/again4you)
45 * [Wook Song](https://github.com/wooksong)
46 * [Jaeyun Jung](https://github.com/jaeyun-jung)
47 * [Hyoungjoo Ahn](https://github.com/helloahn)
48 * [Parichay Kapoor](https://github.com/kparichay)
49 * [Dongju Chae](https://github.com/dongju-chae)
50 * [Gichan Jang](https://github.com/gichan-jang)
51 * [Yongjoo Ahn](https://github.com/anyj0527)
52 * [Jihoon Lee](https://github.com/zhoonit)
53 * [Hyeonseok Lee](https://github.com/lhs8928)
54 * [Mete Ozay](https://github.com/meteozay)
60 This component defines layers which consist of a neural network model. Layers have their own properties to be set.
62 | Keyword | Layer Class Name | Description |
63 |:-------:|:---:|:---|
64 | conv1d | Conv1DLayer | Convolution 1-Dimentional Layer |
65 | conv2d | Conv2DLayer |Convolution 2-Dimentional Layer |
66 | pooling2d | Pooling2DLayer |Pooling 2-Dimentional Layer. Support average / max / global average / global max pooling |
67 | flatten | FlattenLayer | Flatten layer |
68 | fully_connected | FullyConnectedLayer | Fully connected layer |
69 | input | InputLayer | Input Layer. This is not always required. |
70 | batch_normalization | BatchNormalizationLayer | Batch normalization layer |
71 | activation | ActivaitonLayer | Set by layer property |
72 | addition | AdditionLayer | Add input input layers |
73 | attention | AttentionLayer | Attenstion layer |
74 | centroid_knn | CentroidKNN | Centroid K-nearest neighbor layer |
75 | concat | ConcatLayer | Concatenate input layers |
76 | multiout | MultiOutLayer | Multi-Output Layer |
77 | backobne_nnstreamer | NNStreamerLayer | Encapsulate NNStreamer layer |
78 | backbone_tflite | TfLiteLayer | Encapsulate tflite as an layer |
79 | permute | PermuteLayer | Permute layer for transpose |
80 | preprocess_flip | PreprocessFlipLayer | Preprocess random flip layer |
81 | preprocess_l2norm | PreprocessL2NormLayer | Preprocess simple l2norm layer to normalize |
82 | preprocess_translate | PreprocessTranslateLayer | Preprocess translate layer |
83 | reshape | ReshapeLayer | Reshape tensor dimension layer |
84 | split | SplitLayer | Split layer |
85 | dropout | DropOutLayer | Dropout Layer |
86 | embedding | EmbeddingLayer | Embedding Layer |
87 | rnn | RNNLayer | Recurrent Layer |
88 | gru | GRULayer | Gated Recurrent Unit Layer |
89 | lstm | LSTMLayer | Long Short-Term Memory Layer |
90 | lstmcell | LSTMCellLayer | Long Short-Term Memory Cell Layer |
91 | time_dist | TimeDistLayer | Time distributed Layer |
93 ### Supported Optimizers
97 | Keyword | Optimizer Name | Description |
98 |:-------:|:---:|:---:|
99 | sgd | Stochastic Gradient Decent | - |
100 | adam | Adaptive Moment Estimation | - |
102 ### Supported Loss Functions
106 | Keyword | Class Name | Description |
107 |:-------:|:---:|:---:|
108 | cross_sigmoid | CrossEntropySigmoidLossLayer | Cross entropy sigmoid loss layer |
109 | cross_softmax | CrossEntropySoftmaxLossLayer | Cross entropy softmax loss layer |
110 | constant_derivative | ConstantDerivativeLossLayer | Constant derivative loss layer |
111 | mse | MSELossLayer | Mean square error loss layer |
113 ### Supported Activation Functions
117 | Keyword | Loss Name | Description |
118 |:-------:|:---:|:---|
119 | tanh | tanh function | set as layer property |
120 | sigmoid | sigmoid function | set as layer property |
121 | relu | relu function | set as layer propery |
122 | softmax | softmax function | set as layer propery |
123 | weight_initializer | Weight Initialization | Xavier(Normal/Uniform), LeCun(Normal/Uniform), HE(Normal/Unifor) |
124 | weight_regularizer | weight decay ( L2Norm only ) | needs set weight_regularizer_param & type |
125 | learnig_rate_decay | learning rate decay | need to set step |
129 Tensor is responsible for calculation of a layer. It executes several operations such as addition, division, multiplication, dot production, data averaging and so on. In order to accelerate calculation speed, CBLAS (C-Basic Linear Algebra: CPU) and CUBLAS (CUDA: Basic Linear Algebra) for PC (Especially NVIDIA GPU) are implemented for some of the operations. Later, these calculations will be optimized.
130 Currently, we supports lazy calculation mode to reduce complexity for copying tensors during calculations.
132 | Keyword | Description |
134 | 4D Tensor | B, C, H, W|
135 | Add/sub/mul/div | - |
136 | sum, average, argmax | - |
137 | Dot, Transpose | - |
138 | normalization, standardization | - |
145 | Keyword | Loss Name | Description |
146 |:-------:|:---:|:---|
147 | weight_initializer | Weight Initialization | Xavier(Normal/Uniform), LeCun(Normal/Uniform), HE(Normal/Unifor) |
148 | weight_regularizer | weight decay ( L2Norm only ) | needs set weight_regularizer_constant & type |
149 | learnig_rate_decay | learning rate decay | need to set step |
152 Currently, we provide [C APIs](https://github.com/nnstreamer/nntrainer/blob/master/api/capi/include/nntrainer.h) for Tizen. C++ API will be also provided soon.
155 ### Examples for NNTrainer
157 #### [Custom Shortcut Application](https://github.com/nnstreamer/nntrainer/tree/main/Applications/Tizen_native/CustomShortcut)
160 A demo application which enable user defined custom shortcut on galaxy watch.
162 #### [MNIST Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/MNIST)
164 An example to train mnist dataset. It consists two convolution 2d layer, 2 pooling 2d layer, flatten layer and fully connected layer.
166 #### [Reinforcement Learning Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/ReinforcementLearning/DeepQ)
168 A reinforcement learning example with cartpole game. It is using DeepQ algorithm.
170 #### [Transfer Learning Examples](https://github.com/nnstreamer/nntrainer/tree/main/Applications/TransferLearning)
172 Transfer learning examples with for image classification using the Cifar 10 dataset and for OCR. TFlite is used for feature extractor and modify last layer (fully connected layer) of network.
174 #### [ResNet Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/Resnet)
176 An example to train resnet18 network.
178 #### [VGG Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/VGG)
180 An example to train vgg16 network.
182 #### [ProductRating Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/ProductRatings)
184 This application contains a simple embedding-based model that predicts ratings given a user and a product.
186 #### [SimpleShot Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/SimpleShot)
188 An example to demonstrate few-shot learning : SimpleShot
190 #### [Custom Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/Custom)
192 An example to demonstrate how to create custom layers, optimizers or other supported objects.
194 #### ~Tizen CAPI Example~
196 An example to demonstrate c api for Tizen. It is same transfer learing but written with tizen c api.~
197 Deleted instead moved to a [test](https://github.com/nnstreamer/nntrainer/blob/master/test/tizen_capi/unittest_tizen_capi.cpp)
199 #### [KNN Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/KNN)
201 A transfer learning example with for image classification using the Cifar 10 dataset. TFlite is used for feature extractor and compared with KNN.
203 #### [Logistic Regression Example](https://github.com/nnstreamer/nntrainer/tree/main/Applications/LogisticRegression)
205 A logistic regression example using NNTrainer.
207 ## [Getting Started](https://github.com/nnstreamer/nntrainer/blob/main/docs/getting-started.md)
209 Instructions for installing NNTrainer.
211 ### [Running Examples](https://github.com/nnstreamer/nntrainer/blob/main/docs/how-to-run-examples.md)
213 Instructions for preparing NNTrainer for execution
215 ## Open Source License
217 The nntrainer is an open source project released under the terms of the Apache License version 2.0.
221 Contributions are welcome! Please see our [Contributing](https://github.com/nnstreamer/nntrainer/blob/main/docs/contributing.md) Guide for more details.
223 [![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/0)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/0)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/1)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/1)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/2)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/2)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/3)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/3)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/4)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/4)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/5)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/5)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/6)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/6)[![](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/images/7)](https://sourcerer.io/fame/dongju-chae/nnstreamer/nntrainer/links/7)