Merge branch 'v1.2.1_beta'

.Doxyfile

@@ -250,7 +250,7 @@ OPTIMIZE_OUTPUT_FOR_C  = YES
 # qualified scopes will look different, etc.
 # The default value is: NO.
 
-OPTIMIZE_OUTPUT_JAVA   = NO
+OPTIMIZE_OUTPUT_JAVA   = YES
 
 # Set the OPTIMIZE_FOR_FORTRAN tag to YES if your project consists of Fortran
 # sources. Doxygen will then generate output that is tailored for Fortran.
@@ -281,7 +281,7 @@ OPTIMIZE_OUTPUT_VHDL   = NO
 # Note that for custom extensions you also need to set FILE_PATTERNS otherwise
 # the files are not read by doxygen.
 
-EXTENSION_MAPPING      =
+EXTENSION_MAPPING      = proto=C++
 
 # If the MARKDOWN_SUPPORT tag is enabled then doxygen pre-processes all comments
 # according to the Markdown format, which allows for more readable
@@ -771,8 +771,8 @@ WARN_LOGFILE           =
 # spaces. See also FILE_PATTERNS and EXTENSION_MAPPING
 # Note: If this tag is empty the current directory is searched.
 
-INPUT                  = ./inference/engine/api/java ./inference/engine/api/c \
-./inference/flow/include/flow.h ./common/uni/include/task.h ./inference/flow/src/flow.proto
+INPUT                  = ./inference/engine/api ./inference/flow/include/flow.h \
+./common/uni/include/task.h ./inference/flow/src/flow.proto
 
 # This tag can be used to specify the character encoding of the source files
 # that doxygen parses. Internally doxygen uses the UTF-8 encoding. Doxygen uses

.github/workflows/linux-android-armv7.yml

@@ -18,4 +18,5 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           cd ..
+          chmod +x install.sh
           ./install.sh --target=android-armv7 -t 32

.github/workflows/linux-android-armv8.yml

@@ -18,4 +18,5 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           cd ..
-          ./install.sh --target=android-aarch64 --mali
+          chmod +x install.sh
+          ./install.sh --target=android-aarch64

.github/workflows/linux-android-x86_64.yml

@@ -18,4 +18,5 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/linux-x86_64/bin:$PATH
           cd ..
+          chmod +x install.sh
           ./install.sh --target=android-x86_64

.github/workflows/linux-x86-avx2.yml

@@ -12,4 +12,5 @@ jobs:
         uses: actions/checkout@v2
       - name: compile x86-avx2 on linux 
         run: |
+          chmod +x install.sh
           ./install.sh --target=linux-x86_64_avx2 -t 32

.github/workflows/linux-x86.yml

@@ -12,4 +12,5 @@ jobs:
         uses: actions/checkout@v2
       - name: compile x86 on linux
         run: |
+          chmod +x install.sh
           ./install.sh --target=linux-x86_64 -t 32

.github/workflows/macos-android-armv7.yml

@@ -19,6 +19,7 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/darwin-x86_64/bin:$PATH
           cd ..
+          chmod +x install.sh
           ./install.sh --target=android-armv7 
           
           

.github/workflows/macos-android-armv8.yml

@@ -19,4 +19,5 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/darwin-x86_64/bin:$PATH
           cd ..
-          ./install.sh --target=android-aarch64 --mali
+          chmod +x install.sh
+          ./install.sh --target=android-aarch64 

.github/workflows/macos-android-x86_64.yml

@@ -19,4 +19,5 @@ jobs:
           export ANDROID_NDK_ROOT=${PWD}/android-ndk-r20b
           export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/darwin-x86_64/bin:$PATH
           cd ..
+          chmod +x install.sh
           ./install.sh --target=android-x86_64

.github/workflows/macos-ios-armv7.yml

@@ -14,4 +14,5 @@ jobs:
           echo 'export FLAGS_GETOPT_CMD="$(brew --prefix gnu-getopt)/bin/getopt"' >> /Users/runner/.bash_profile
           source /Users/runner/.bash_profile        
           export IOS_SDK_ROOT=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk
+          chmod +x install.sh
           ./install.sh --target=ios-armv7 

.github/workflows/macos-ios-armv8.yml

@@ -14,4 +14,5 @@ jobs:
           echo 'export FLAGS_GETOPT_CMD="$(brew --prefix gnu-getopt)/bin/getopt"' >> /Users/runner/.bash_profile
           source /Users/runner/.bash_profile        
           export IOS_SDK_ROOT=/Applications/Xcode.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS.sdk
+          chmod +x install.sh
           ./install.sh --target=ios-aarch64 

.github/workflows/macos-x86-avx2.yml

@@ -21,4 +21,5 @@ jobs:
           gcc -v
           g++ -v
           cd $BOLT_PATH
+          chmod +x install.sh
           ./install.sh --target=macos-x86_64_avx2 -t 32

.github/workflows/macos-x86.yml

@@ -13,4 +13,4 @@ jobs:
           echo 'export PATH="/usr/local/opt/gnu-getopt/bin:$PATH"' >> /Users/runner/.bash_profile
           echo 'export FLAGS_GETOPT_CMD="$(brew --prefix gnu-getopt)/bin/getopt"' >> /Users/runner/.bash_profile
           source /Users/runner/.bash_profile
-          ./install.sh --target=macos-x86_64
+          chmod +x install.sh && ./install.sh --target=macos-x86_64

.github/workflows/windows-android-armv8.yml

@@ -33,5 +33,5 @@ jobs:
         export PATH=${ANDROID_NDK_ROOT}/toolchains/llvm/prebuilt/windows-x86_64/bin:$PATH
         export PATH=$PATH:/d/a/bolt/bolt/software/wget
         cd /d/a/bolt/bolt
-        ./install.sh --target=android-aarch64 --mali
+        ./install.sh --target=android-aarch64 
       shell: bash

.gitignore

@@ -6,15 +6,59 @@ lib/*
 */install_*/*
 */bin/*
 */lib/*
-
-third_party/*
-
-caffe_pb2.py
 *.o
 *.a
 *.so
+*.bin
+*.swp
+
+_book
+node_modules
+__pycache__
+inference/flow/include/flow.pb.h
+caffe_pb2.py
+
+model_tools/tools/tensorflow2caffe/*/*.txt
+model_tools/tools/tensorflow2caffe/*/*.npy
+
 *.bolt
+*.onnx
+*.tflite
+*.pb
+*.prototxt
+*.caffemodel
+*.pdf
+
+third_party
+
+common/gcl/tools/kernel_lib_compile/bin2char
+common/gcl/tools/kernel_lib_compile/gcl_binary
+common/gcl/tools/kernel_lib_compile/gcl_device_name
+common/gcl/tools/kernel_lib_compile/include
+common/gcl/tools/kernel_lib_compile/src
+common/gcl/tools/kernel_lib_compile/build
+common/gcl/tools/kernel_lib_compile/name
+common/gcl/tools/kernel_lib_compile/cl
+common/gcl/kernelBin
+common/gcl/tools/kernel_source_compile/include
+common/gcl/tools/kernel_source_compile/src
+kit/iOS/ImageClassification/ImageClassificationDemo/libbolt/imagenet_classes.txt
+kit/iOS/ImageClassification/ImageClassificationDemo/libbolt/headers
+kit/Android/ImageClassification/app/src/main/assets/imagenet_classes.txt
+kit/Android/ImageClassification/app/src/main/cpp/libbolt/headers
+kit/Android/CameraEnlarge/app/src/main/cpp/libbolt/headers
+kit/iOS/CameraEnlarge/CameraEnlarge/libbolt/headers
+kit/Android/SimpleImageClassification/app/src/main/assets
+kit/Android/SimpleImageClassification/app/src/main/java
+kit/iOS/SimpleImgClassfication/libbolt
+kit/Android/Semantics/app/src/main/java
+kit/Android/Semantics/app/src/main/assets/
+
 
-docs/API/html
-kit/Android
-kit/iOS
+final_combinations.txt
+single_combinations.txt
+mt_result.txt
+ptq_result.txt
+engine_result.txt
+tmp.txt
+report.csv

CMakeLists.txt

@@ -16,9 +16,9 @@ endif (USE_TFLITE)
 if (USE_TENSORFLOW)
     find_package(jsoncpp)
 endif (USE_TENSORFLOW)
-if (USE_MALI)
+if (USE_GPU)
     find_package(Gcl)
-endif(USE_MALI)
+endif(USE_GPU)
 if (NOT "$ENV{JNI_ROOT}" STREQUAL "")
     find_package(JNI)
     if (JNI_FOUND)
@@ -27,6 +27,9 @@ if (NOT "$ENV{JNI_ROOT}" STREQUAL "")
 endif ()
 if (BUILD_TEST)
     find_package(jpeg)
+    if (EXISTS ${OpenCV_CMAKE_PATH})
+        include(${OpenCV_CMAKE_PATH})
+    endif()
 endif (BUILD_TEST)
 
 add_subdirectory(common)
@@ -75,9 +78,9 @@ if (BASH_PROGRAM AND USE_GENERAL)
     else()
         set(parameters ${parameters} -d unknown)
     endif()
-    if (USE_MALI)
+    if (USE_GPU)
         set(parameters ${parameters} -g)
-    endif(USE_MALI)
+    endif(USE_GPU)
     if (USE_DYNAMIC_LIBRARY)
         set(parameters ${parameters} --shared)
     endif(USE_DYNAMIC_LIBRARY)

README.md

@@ -1,17 +1,17 @@
 # Introduction
-
+---
 [![License: MIT](docs/images/license-mit.png)](https://opensource.org/licenses/MIT)
 
 [Bolt](https://huawei-noah.github.io/bolt/) is a light-weight library for deep learning. Bolt, as a universal deployment tool for all kinds of neural networks, aims to minimize the inference runtime as much as possible.
 Bolt has been widely deployed and used in many departments of HUAWEI company, such as 2012 Laboratory, CBG and HUAWEI Product Lines.
 If you have questions or suggestions, you can submit issue. **QQ群: 833345709**
 
 # Why Bolt is what you need?
-
+---
 - **High Performance:** **15%+** faster than existing open source acceleration libraries.
 - **Rich Model Conversion:** support Caffe, ONNX, TFLite, Tensorflow.
 - **Various Inference Precision:** support FP32, FP16, INT8, 1-BIT.
-- **Multiple platforms:** ARM CPU(v7, v8, v8.2), Mali GPU, X86(AVX2)
+- **Multiple platforms:** ARM CPU(v7, v8, v8.2), Mali GPU, Qualcomm GPU, X86 CPU(AVX2, AVX512)
 - **Bolt is the first to support NLP and also supports common CV applications.**
 - **Minimize ROM/RAM**
 - Rich Graph Optimization
@@ -22,13 +22,13 @@ If you have questions or suggestions, you can submit issue. **QQ群: 833345709**
 [See more excellent features and details here](https://zhuanlan.zhihu.com/p/317111024)
 
 # Building Status
-
+---
 Kinds of choices are provided for the compilation of bolt. Please make a suitable choice depending on your environment.
 
 | target platform      | build command                                | Linux | Windows | MacOS |
 | -------------------- | -------------------------------------------- | ----- | ------- | ----- |
 | Android(armv7)       | ./install.sh --target=android-armv7          | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/linux-android-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Alinux-android-armv7) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/windows-android-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Awindows-android-armv7) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-android-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-android-armv7) |
-| Android(armv8+mali)  | ./install.sh --target=android-aarch64 --mali | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/linux-android-armv8)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Alinux-android-armv8) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/windows-android-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Awindows-android-armv8) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-android-armv8)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-android-armv8) |
+| Android(armv8+gpu)  | ./install.sh --target=android-aarch64 --gpu | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/linux-android-armv8)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Alinux-android-armv8) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/windows-android-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Awindows-android-armv8) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-android-armv8)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-android-armv8) |
 | Android(x86_64)      | ./install.sh --target=android-x86_64         | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/linux-android-x86_64)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Alinux-android-x86_64) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/windows-android-x86_64)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Awindows-android-x86_64) | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-android-x86_64)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-android-x86_64) |
 | iOS(armv7)           | ./install.sh --target=ios-armv7              | / | / | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-ios-armv7)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-ios-armv7) |
 | iOS(armv8)           | ./install.sh --target=ios-aarch64            | / | / | [![Build Status](https://img.shields.io/github/workflow/status/huawei-noah/bolt/macos-ios-armv8)](https://github.com/huawei-noah/bolt/actions?query=workflow%3Amacos-ios-armv8) |
@@ -42,19 +42,25 @@ Kinds of choices are provided for the compilation of bolt. Please make a suitabl
 *NOTE: Bolt defaultly link static library, This may cause some problem on some platforms. You can use --shared option to link shared library.*
 
 # Quick Start
-
-<div align=center><img src="docs/images/QuickStart.jpg" width = 80% height = 80% /></div>
-
+---
+<div align=center><img src="docs/images/QuickStart.jpg" width = 100% height = 100%  style="border: 1px solid rgba(151,151,151,0.50)" /></div>
 Two steps to get started with bolt.
 
-1. Conversion: use **[X2bolt](/model_tools/tools/X2bolt/X2bolt.cpp)** to convert your model from caffe,onnx,tflite or tensorflow to .bolt;
+1. Conversion: use **[X2bolt](model_tools/tools/X2bolt/X2bolt.cpp)** to convert your model from caffe,onnx,tflite or tensorflow to .bolt;
 
-2. Inference: run **[benchmark](/inference/examples/benchmark/benchmark.cpp)** with .bolt and data to get the inference result.
+2. Inference: run **[benchmark](inference/examples/benchmark/benchmark.cpp)** with .bolt and data to get the inference result.
 
    For more details about the usage of [**X2bolt**](model_tools/tools/X2bolt/X2bolt.cpp) and [**benchmark**](inference/examples/benchmark/benchmark.cpp) tools,  see [docs/USER_HANDBOOK.md](docs/USER_HANDBOOK.md). 
 
-# Verified Networks
+# DL Applications in Bolt
+
+  Here we show some interesting and useful applications in bolt.
 
+| Face Detection | ASR | Semantics Analysis | Image Classification
+| :------: | :------: | :------: |:------: 
+| [![face_detection](docs/images/20_bolt_face_detection.gif)](inference/examples/ultra_face) demo_link: [face detection](inference/examples/ultra_face) | [![asr](docs/images/ChineseSpeechRecognition.gif)]()  demo_link: [asr](inference/examples/automatic_speech_recognition) | [![semantics analysis](docs/images/SemanticsAnalysis.gif)]()  demo_link: [semantics analysis](kit/Android/Semantics) | [![image_classification](docs/images/ImageClassification.gif)]() demo_link: [image classification](inference/examples/image_classification) 
+# Verified Networks
+---
   Bolt has shown its high performance in the inference of common CV and NLP neural networks. Some of the representative networks that we have verified are listed below. You can find detailed benchmark information in [docs/BENCHMARK.md](docs/BENCHMARK.md).
 
   <table border="1"  bordercolor="#00CCCC"  width="300">
@@ -75,7 +81,7 @@ Two steps to get started with bolt.
         <a href="https://github.com/liuzechun/Bi-Real-net">BiRealNet</a>,
         <a href="https://github.com/liuzechun/ReActNet">ReActNet</a>,
         <a href="https://github.com/huawei-noah/ghostnet">Ghostnet</a>,
-        SSD, Yolov3, Pointnet ...
+        SSD, Yolov3, Pointnet, ViT, TNT ...
       </td>
     </tr>
     <tr>
@@ -87,6 +93,14 @@ Two steps to get started with bolt.
         <a href="docs/USER_HANDBOOK.md#voice-wake-up">Tdnn</a> ...
       </td>
     </tr>
+    <tr>
+      <td>
+        Recommendation
+      </td>
+      <td>
+        MLP
+      </td>
+    </tr>
     <tr>
       <td>
         More DL Tasks
@@ -100,7 +114,7 @@ Two steps to get started with bolt.
   More models than these mentioned above are supported,  users are encouraged to further explore.
 
 # Documentations
-
+---
 Everything you want to know about bolt is recorded in the detailed documentations stored in [docs](docs).
 
 - [How to install bolt with different compilers](docs/INSTALL.md).
@@ -112,7 +126,7 @@ Everything you want to know about bolt is recorded in the detailed documentation
 - [Frequently Asked Questions(FAQ)](docs/FAQ.md)
 
 # Articles
-
+---
 - [深度学习加速库Bolt领跑端侧AI](https://zhuanlan.zhihu.com/p/317111024)
 - [为什么 Bolt 这么快：矩阵向量乘的实现](https://zhuanlan.zhihu.com/p/97928435)
 - [深入硬件特性加速TinyBert，首次实现手机上Bert 6ms推理](https://zhuanlan.zhihu.com/p/158620259)
@@ -121,13 +135,17 @@ Everything you want to know about bolt is recorded in the detailed documentation
 - [ARM CPU 1-bit推理，走向极致的道路](https://zhuanlan.zhihu.com/p/158161592)
 
 # 教程
-
+---
 - 图像分类: [Android Demo](https://zhuanlan.zhihu.com/p/359264809), [iOS Demo](https://zhuanlan.zhihu.com/p/359259766)
+- 图像增强: [Android Deme](), [iOS Demo]()
+- 情感分类: [Android Demo](https://zhuanlan.zhihu.com/p/414971037)
+- 中文语音识别: [Android Demo](https://zhuanlan.zhihu.com/p/414978782), [iOS Demo](https://zhuanlan.zhihu.com/p/414981121)
+- 人脸检测: [Android Demo](https://zhuanlan.zhihu.com/p/414975102), [iOS Demo](https://zhuanlan.zhihu.com/p/414971375)
 
 # Acknowledgement
-
+---
 Bolt refers to the following projects: [caffe](https://github.com/BVLC/caffe), [onnx](https://github.com/onnx/onnx), [tensorflow](https://github.com/tensorflow/tensorflow), [ncnn](https://github.com/Tencent/ncnn), [mnn](https://github.com/alibaba/MNN), [dabnn](https://github.com/JDAI-CV/dabnn).
 
 # License
-
+---
 The MIT License(MIT)

common/CMakeLists.txt

@@ -13,7 +13,7 @@ endif (BOLT_CONFIGURE_FILE)
 project(common)
 
 add_subdirectory(uni)
-if (USE_MALI)
+if (USE_GPU)
     add_subdirectory(gcl)
-endif (USE_MALI)
+endif (USE_GPU)
 add_subdirectory(model_spec)

common/cmakes/FindProtobuf.cmake

@@ -9,7 +9,7 @@ if (USE_DYNAMIC_LIBRARY)
     set(Protobuf_SHARED_LIBRARY ${Protobuf_LIBRARY})
 else (USE_DYNAMIC_LIBRARY)
     find_library(Protobuf_LIBRARY NAMES ${CMAKE_STATIC_LIBRARY_PREFIX}protobuf${CMAKE_STATIC_LIBRARY_SUFFIX} HINTS $ENV{Protobuf_ROOT}/lib ${Protobuf_ROOT}/lib)
-    find_library(Protobuf_SHARED_LIBRARY NAMES protobuf SHARED_HINTS $ENV{Protobuf_ROOT}/lib ${Protobuf_ROOT}/lib)
+    find_library(Protobuf_SHARED_LIBRARY NAMES protobuf HINTS $ENV{Protobuf_ROOT}/lib ${Protobuf_ROOT}/lib)
 endif (USE_DYNAMIC_LIBRARY)
 
 find_program(Protobuf_PROTOC_EXECUTABLE NAMES protoc HINTS $ENV{Protobuf_ROOT}/bin ${Protobuf_ROOT}/bin)