Ready for v4.2.0.4 (#6)

* experiments with mkml

* Update README.md

* Update README.md

* openblas experiments

* Update README.md

* Update README.md

* Update README.md

* Update README.md

* intel license

* no IPP

* Update README.md

* openblas right way

* openblas no gfortran

* openblas win

* rm submodules

* dnld fix

* minor fixes

* :# This is a combination of 4 commits.

README upd

Update README.md

Update README.md

Update README.md

Author: banderlog

.gitignore

@@ -1,30 +1,47 @@
-*cache*
-.ipynb_checkpoints
+openblas/*
+!openblas/.gitkeep
+
+dldt/*
+!dldt/.gitkeep
+
+opencv/*
+!opencv/.gitkeep
+
+ffmpeg/*
+!ffmpeg/.gitkeep
+
 build/*
 !build/opencv
 build/opencv/*
 !build/opencv/opencv_setup.sh
+
 !build/dldt
 build/dldt/*
 !build/dldt/dldt_setup.sh
+
 !build/ffmpeg
 build/ffmpeg/*
 !build/ffmpeg/ffmpeg_setup.sh
 !build/ffmpeg/ffmpeg_premake.sh
+
+!build/openblas
+build/openblas/*
+!build/openblas/openblas_setup.sh
+
 create_wheel/*
 !create_wheel/LICENSE*
 !create_wheel/README.md
 !create_wheel/setup.py
-create_wheel/cv2/*
 !create_wheel/cv2
+create_wheel/cv2/*
 !create_wheel/cv2/__init__.py
-dldt/*
-opencv/*
-ffmpeg/*
-!ffmpeg/.gitkeep
-venv
+
 tests/venv_t
 tests/rateme*
+venv
+
+*cache*
+.ipynb_checkpoints
 *tar.gz
 *tar.bz2
 *.zip

README.md

@@ -2,30 +2,7 @@
 
 # opencv-python-inference-engine
 
-It is *Unofficial* pre-built OpenCV+dldt_module package for Python.
-
-**Why:**  
-There is a [guy with an exellent pre-built set of OpenCV packages](https://github.com/skvark/opencv-python), but they are all came without [dldt module](https://github.com/opencv/dldt). And you need that module if you want to run models from [Intel's model zoo](https://github.com/opencv/open_model_zoo/).
-
-**UPD:** Presumably, you could also use speech2text model now -- [source](https://docs.openvinotoolkit.org/latest/_inference_engine_samples_speech_libs_and_demos_Speech_libs_and_demos.html)
-
-**Limitations**:
-+ Package comes without contrib modules.
-+ You need to [add udev rules](https://github.com/opencv/dldt/blob/2019/inference-engine/README.md#for-linux-raspbian-stretch-os) if you want working MYRIAD plugin.
-+ It was tested on Ubuntu 18.04, Ubuntu 18.10 as Windows 10 Subsystem and Gentoo.
-+ It will not work for Ubuntu 16.04 and below (except v4.1.0.4).
-+ I had not made builds for Windows or MacOS.
-+ It is 64 bit.
-+ It built with `ffmpeg` and `v4l` support (`ffmpeg` libs included).
-+ No GTK/QT support -- use `matplotlib` for plotting your results.
-
-This package is most similar to `opencv-python-headless`, main differences are:
-+ Usage of `AVX2` instructions
-+ No `JPEG 2000`, `WEBP`, `OpenEXR` support
-+ `TBB` used as a parallel framework
-+ Inference Engine with `MYRIAD` plugin
-
-For additional info read `cv2.getBuildInformation()` output.
+This is *Unofficial* pre-built OpenCV with the inference engine part of [dldt module](https://github.com/opencv/dldt/) package for Python.
 
 ## Installing from `pip3`
 
@@ -35,109 +12,109 @@ Remove previously installed versions of `cv2`
 pip3 install opencv-python-inference-engine
 ```
 
-## Downloading intel models
-
-As it appeared it is not an obvious task :)
 
-Please refer to: https://github.com/opencv/open_model_zoo/issues/522
+## Why  
 
-## Known problems and TODOs
+I needed an ability to fast deploy a small package that able to run models from [Intel's model zoo](https://github.com/opencv/open_model_zoo/) and use [Movidius NCS](https://software.intel.com/en-us/neural-compute-stick).
+Wellr-known [opencv-python](https://github.com/skvark/opencv-python) can't do this.
+The official way is to use OpenVINO, but it is big and clumsy (just try to use it with python venv or fast download it on cloud instance).
 
-### No Ubuntu 16.04 support
 
-Release [v4.1.0.4](https://github.com/banderlog/opencv-python-inference-engine/releases/tag/v4.1.0.4) is working with Ubuntu 16.04 ([#3](https://github.com/banderlog/opencv-python-inference-engine/issues/3)).
+## Description
 
-All releases before it were compiled on Ubuntu 18.04 and it has different versions of `glibc`, `cmake`, etc.
-So make it Ubuntu 16.04 compatible was pretty easy -- just change one standard build environment to another.
+### Limitations
 
-But `dldt-2019R2` requires `cmake-3.7.2`, which is absent in Ubuntu 16.04. And, of course, it can be installed, but works-from-the-box behavior is loosed 
-from now on.
-
-### No GTK/QT support
++ Package comes without contrib modules.
++ You need to [add udev rules](https://github.com/opencv/dldt/blob/2019/inference-engine/README.md#for-linux-raspbian-stretch-os) if you want working MYRIAD plugin.
++ It was tested on Ubuntu 18.04, Ubuntu 18.10 as Windows 10 Subsystem and Gentoo.
++ It will not work for Ubuntu 16.04 and below (except v4.1.0.4).
++ I had not made builds for Windows or MacOS.
++ It built with `ffmpeg` and `v4l` support (`ffmpeg` libs included).
++ No GTK/QT support -- use `matplotlib` for plotting your results.
++ It is 64 bit.
 
-[skvarks's package](https://github.com/skvark/opencv-python) has `Qt4` GUI for `opencv` and it is +16 MB to file size.
-Also it is a lot of problems and extra work to compile Qt\GTK libs from sources.
-In 95% of cases `matplotlib.imshow()` will be sufficient, in other 5% use another package for now or compile it with GUI
-support by yourself.
+### Main differences from `opencv-python-headless`
 
-#### Steps to compile it with `GTK-2` support (checked)
++ Usage of `AVX2` instructions
++ No `JPEG 2000`, `WEBP`, `OpenEXR` support
++ `TBB` used as a parallel framework
++ Inference Engine with `MYRIAD` plugin
 
-Make next changes in `opencv-python-inference-engine/build/opencv/opencv_setup.sh`:
-1. change string `-D WITH_GTK=OFF \`  to `-D WITH_GTK=ON \`
-2. change `export PKG_CONFIG_PATH=$ABS_PORTION/build/ffmpeg/binaries/lib/pkgconfig:$PKG_CONFIG_PATH` -- you will need to
-   add absolute paths to `.pc` files. On Ubuntu 18.04 it were
-   `/usr/lib/x86_64-linux-gnu/pkgconfig/:/usr/share/pkgconfig/:/usr/local/lib/pkgconfig/:/usr/lib/pkgconfig/`
+### Main differences from OpenVINO
 
-Exporting `PKG_CONFIG_PATH` for `ffmpeg` somehow messes with default values.
++ No model-optimizer
++ OpenBLAS instead of MKL ([details](https://github.com/banderlog/opencv-python-inference-engine/issues))
++ No [ITT](https://software.intel.com/en-us/articles/intel-itt-api-open-source)
++ No [IPP](https://software.intel.com/en-us/ipp)
++ No [Intel Media SDK](https://software.intel.com/en-us/media-sdk)
++ No [OpenVINO IE API](https://github.com/opencv/dldt/tree/2020/inference-engine/ie_bridges/python/src/openvino/inference_engine)
++ No python2 support (it is dead)
++ No Gstreamer (use ffmpeg)
++ No GTK (+16 MB and a lot of problems and extra work to compile Qt\GTK libs from sources.)
 
-### Not really `manylinux1`
+For additional info read `cv2.getBuildInformation()` output.
 
-The package is renamed to `manylinux1` from `linux`, because, according to [PEP 513](https://www.python.org/dev/peps/pep-0513/), PyPi repo does not want to apply other architectures.
-And compiling it for CentOS 5.11 is pretty challenging (there is no such lxd container plus I do not want to mess with docker) and denies from using some of the necessary libs (like tbb).
-Also, I suspect that it will be poorly optimized.
+### Versioning
 
-### Build `ffmpeg` with `tbb`
+The first 3 letters are the version of OpenCV, the last one -- package version. E.g, `4.1.0.2` -- 2nd version of based on 4.1.0 OpenCV package. Package versions are not continuously numbered -- each new OpenCV version starts its own numbering.
 
-Both `dldt` and `opencv` are compiled with `tbb` support, and `ffmpeg` compiled without it -- this does not feels right.
-There is some unproved solution for how to compile `ffmpeg` with `tbb` support:
-+ <https://stackoverflow.com/questions/6049798/ffmpeg-mt-and-tbb>  
-+ <https://stackoverflow.com/questions/14082360/pthread-vs-intel-tbb-and-their-relation-to-openmp>
 
-Maybe someday I will try it.
+## Downloading intel models
 
-### Versioning
+The official way is clumsy because you need to git clone the whole [model_zoo](https://github.com/opencv/open_model_zoo) ([details](https://github.com/opencv/open_model_zoo/issues/522))
 
-First 3 letters are the version of OpenCV, the last one -- package version. E.g, `4.1.0.2` -- 2nd version of based on 4.1.0 OpenCV package. Package versions are not continuously numbered -- each new OpenCV version starts its own numbering.
+Better to find a model description [here](https://github.com/opencv/open_model_zoo/blob/master/models/intel/index.md) and download manually from [here](https://download.01.org/opencv/2020/openvinotoolkit/2020.1/open_model_zoo/models_bin/1/)
 
 
 ## Compiling from source
 
-I compiled it on Ubuntu 18.04 Linux Container.
+You will need ~6GB RAM and ~10GB disk space
+
+I am using Ubuntu 18.04 [multipass](https://multipass.run/) instance: `multipass launch -c 6 -d 10G -m 7G`.
 
 ### Requirements
 
-+ <https://github.com/opencv/dldt/blob/2018/inference-engine/README.md> 
-+ <https://docs.opencv.org/4.0.0/d7/d9f/tutorial_linux_install.html> (`build-essential`, `cmake`, `git`, `pkg-config`, `python3-dev`)
-+ `nasm` (for ffmpeg)
-+ `python3`
++ `build-essential`
++ `cmake`
++ `git`
++ `pkg-config`
++ `python3-dev`
 + `virtualenv`
-+ `libusb-1.0-0-dev` (for dldt  >= 2019_R1.0.1)
 + `chrpath`
-+ `patchelf`
++ `libusb-1.0-0-dev` (for MYRIAD plugin)
++ `nasm` (for ffmpeg)
 
 ```
 sudo apt-get update
-sudo apt install build-essential cmake git pkg-config python3-dev nasm python3 virtualenv libusb-1.0-0-dev chrpath patchelf
+sudo apt install build-essential cmake git pkg-config python3-dev nasm python3 virtualenv libusb-1.0-0-dev chrpath
 ```
 
-**NB:** you will need ~6GB RAM and ~7GB disk space
-
 ### Preparing
 
-0. `git clone https://github.com/banderlog/opencv-python-inference-engine`
-1. `cd opencv-python-inference-engine`
-2. run `download_all_stuff.sh` (refer for script code for details)
+1. `git clone https://github.com/banderlog/opencv-python-inference-engine`
+2. `cd opencv-python-inference-engine`
+3. run `download_all_stuff.sh` (refer for script code for details)
 
 ### Compilation
 
-`$ABS_PORTION` is a absolute path to `opencv-python-inference-engine` dir.
-
 ```bash
-
-cd build/ffmpeg
-./ffmpeg_setup.sh
-./ffmpeg_premake.sh
-make -j8
+cd build/openblas
+./openblas_setup.sh &&
+make -j8 &&
 make install
 
+cd ../ffmpeg
+./ffmpeg_setup.sh &&
+./ffmpeg_premake.sh &&
+make -j8 &&
+make install
 
 cd ../dldt
-./dldt_setup.sh
+./dldt_setup.sh &&
 make -j8
 
-
 cd ../opencv
-ABS_PORTION=YOUR_ABSOLUTE_PATH_TO_opencv-python-inference-engine_dir ./opencv_setup.sh
+./opencv_setup.sh &&
 make -j8
 ```
 
@@ -155,11 +132,82 @@ cp dldt/inference-engine/temp/tbb/lib/libtbb.so.2 create_wheel/cv2/
 
 cp build/ffmpeg/binaries/lib/*.so create_wheel/cv2/
 
+cp build/openblas/lib/libopenblas.so.0 create_wheel/cv2/
 
-cd create_wheel
 # change RPATH
+cd create_wheel
 for i in  cv2/*.so; do chrpath -r '$ORIGIN' $i; done
 
 # final .whl will be in /create_wheel/dist/
 ../venv/bin/python3 setup.py bdist_wheel
 ```
+
+### Optional things to play with
+
++ [dldt build instruction](https://github.com/opencv/dldt/blob/2020/build-instruction.md)
++ [dldt cmake flags](https://github.com/opencv/dldt/blob/b2140c083a068a63591e8c2e9b5f6b240790519d/inference-engine/cmake/features_ie.cmake)
++ [opencv cmake flags](https://github.com/opencv/opencv/blob/master/CMakeLists.txt)
+
+**NB:** removing `QUIET` from `find_package()` in project Cmake files, could help to solve some problems -- сmake will start to log them.
+
+
+#### GTK2
+
+Make next changes in `opencv-python-inference-engine/build/opencv/opencv_setup.sh`:
+1. change string `-D WITH_GTK=OFF \`  to `-D WITH_GTK=ON \`
+2. `export PKG_CONFIG_PATH=$ABS_PORTION/build/ffmpeg/binaries/lib/pkgconfig:$PKG_CONFIG_PATH` -- you will need to
+   add absolute paths to `.pc` files. On Ubuntu 18.04 they here:
+   `/usr/lib/x86_64-linux-gnu/pkgconfig/:/usr/share/pkgconfig/:/usr/local/lib/pkgconfig/:/usr/lib/pkgconfig/`
+
+Exporting `PKG_CONFIG_PATH` for `ffmpeg` somehow messes with default values.
+
+#### IPP
+
+Just set `-D WITH_IPP=ON` in `opencv_setup.sh`.
+
+It will give +30MB to the final `cv2.so` size. And it will boost _some_ opencv functions.
+
+![](https://www.oreilly.com/library/view/learning-opencv-3/9781491937983/assets/lcv3_0105.png)
+(Image from [Learning OpenCV 3 by Gary Bradski, Adrian Kaehler](https://www.oreilly.com/library/view/learning-opencv-3/9781491937983/ch01.html))
+
+[Official Intel's IPP benchmarks](https://software.intel.com/en-us/ipp/benchmarks) (may ask for registration)
+
+#### MKL
+
+You need to download MKL-DNN release and set two flags:`-D GEMM=MKL` , `-D MKLROOT` ([details](https://github.com/opencv/dldt/issues/327))
+
+OpenVino comes with 30MB `libmkl_tiny_tbb.so`, but [you will not be able to compile it](https://github.com/intel/mkl-dnn/issues/674), because it made from proprietary MKL.
+
+Our opensource MKL-DNN experiment will end with 125MB `libmklml_gnu.so` and inference speed compatible with 35MB openblas ([details](https://github.com/banderlog/opencv-python-inference-engine/issues/5)).
+
+#### OpenBLAS
+
+Please refer here for details: https://github.com/xianyi/OpenBLAS/issues/2528
+
++ [OpenBLAS Installation guide](https://github.com/xianyi/OpenBLAS/wiki/Installation-Guide)
++ [OpenBLAS User Manual](https://github.com/xianyi/OpenBLAS/wiki/User-Manual)
+
+If you compile it with `make FC=gfortran`, you'll need to put `libgfortran.so.4` and `libquadmath.so.0` to wheel and set them rpath via `patchelf --set-rpath \$ORIGIN *.so`
+
+
+#### CUDA
+
+I did not try it.
+
++ [Compile OpenCV’s ‘dnn’ module with NVIDIA GPU support](https://www.pyimagesearch.com/2020/02/10/opencv-dnn-with-nvidia-gpus-1549-faster-yolo-ssd-and-mask-r-cnn/)
++ [Use OpenCV’s ‘dnn’ module with NVIDIA GPUs, CUDA, and cuDNN](https://www.pyimagesearch.com/2020/02/03/how-to-use-opencvs-dnn-module-with-nvidia-gpus-cuda-and-cudnn/)
+
+
+#### Build `ffmpeg` with `tbb`
+
+Both `dldt` and `opencv` are compiled with `tbb` support, and `ffmpeg` compiled without it -- this does not feel right.
+There is some unproved solution for how to compile `ffmpeg` with `tbb` support: <https://stackoverflow.com/questions/6049798/ffmpeg-mt-and-tbb>  
+
+
+#### OpenMP
+
+It is possible to compile OpenBLAS, dldt and OpenCV with OpenMP. I am not sure that the result would be better than now, but who knows.
+
+#### Use opencv for NLP
+
+Presumably, you could also use speech2text model now -- [source](https://docs.openvinotoolkit.org/latest/_inference_engine_samples_speech_libs_and_demos_Speech_libs_and_demos.html)