OpenVINO-YoloV3

YoloV3 / tiny-YoloV3 + RaspberryPi3 / Ubuntu LaptopPC + NCS/NCS2 + USB Camera + Python

Inspired from https://github.com/mystic123/tensorflow-yolo-v3.git

Performance comparison as a mobile application (Based on sensory comparison)
◯=HIGH, △=MEDIUM, ×=LOW

My articles

1. [24 FPS] Boost RaspberryPi3 with four Neural Compute Stick 2 (NCS2) MobileNet-SSD / YoloV3 [48 FPS for Core i7]

2. [13 FPS] NCS2 x 4 + Full size YoloV3 performance has been tripled

3. Support for local training and OpenVINO of One Class tiny-YoloV3 with a proprietary data set

Change history

[Mar 01, 2019]　Improve accuracy. Fixed preprocessing and postprocessing bug.
[Mar 17, 2019]　Added a training procedure with your own data set.
[Apr 03, 2019]　Work on OpenVINO 2019 R1 started.
[Apr 14, 2019]　Compatible with 2019 R1.
[Apr 26, 2019]　Compatible with 2019 R1.0.1.

Operation sample

<CPP + YoloV3 - Intel Core i7-8750H, CPU Only, 4 FPS - 5 FPS>

<CPP + tiny-YoloV3 - Intel Core i7-8750H, CPU Only, 60 FPS>

<Python + tiny-YoloV3 + USBCamera, Core i7-8750H, CPU Only, 30 FPS>

<Python + YoloV3 + MP4, Core i7-8750H, NCS2 x4, 13 FPS>
【Note】 Due to the performance difference of ARM <-> Core series, performance is degraded in RaspberryPi3.

Python Version YoloV3 / tiny-YoloV3 (Dec 28, 2018 Operation confirmed)

YoloV3

$ python3 openvino_yolov3_test.py

tiny-YoloV3 + NCS2 MultiStick

$ python3 openvino_tiny-yolov3_MultiStick_test.py -numncs 1

YoloV3 + NCS2 MultiStick (Pretty slow)

$ python3 openvino_yolov3_MultiStick_test.py -numncs 4

CPP Version YoloV3 / tiny-YoloV3 (Dec 16, 2018 Operation confirmed)

cpp version is here "cpp/object_detection_demo_yolov3_async"

Environment

• LattePanda Alpha (Intel 7th Core m3-7y30) or LaptopPC (Intel 8th Core i7-8750H)

• Ubuntu 16.04 x86_64

• RaspberryPi3

• Raspbian Stretch armv7l

• OpenVINO toolkit 2019 R1.0.1 (2019.1.133)

• Python 3.5

• OpenCV 4.1.0-openvino

• Tensorflow v1.12.0 or Tensorflow-GPU v1.12.0 (pip install)

• YoloV3 (MS-COCO)

• tiny-YoloV3 (MS-COCO)

• USB Camera (PlaystationEye) / Movie file (mp4)

• Intel Neural Compute Stick v1 / v2

OpenVINO Supported Layers (As of Apr 14, 2019)

• Supported Framework Layers

• Supported Caffe Layers

• Supported TensorFlow Layers

• Supported MXNet Layers

• Supported ONNX Layers

Supported Devices (https://docs.openvinotoolkit.org/latest/_docs_IE_DG_supported_plugins_Supported_Devices.html#supported_layers)

OpenVINO - Python API

https://docs.openvinotoolkit.org/latest/_inference_engine_ie_bridges_python_docs_api_overview.html

Environment construction procedure

1. Work with LaptopPC (Ubuntu 16.04)

1.OpenVINO 2019R1.0.1 Full-Install. Execute the following command.

$ cd ~$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1ciX7cHqCh8lLFYI0HKkhC3r_fMirrlKk" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1ciX7cHqCh8lLFYI0HKkhC3r_fMirrlKk" -o l_openvino_toolkit_p_2019.1.133.tgz
$ tar -zxf l_openvino_toolkit_p_2019.1.133.tgz
$ rm l_openvino_toolkit_p_2019.1.133.tgz
$ cd l_openvino_toolkit_p_2019.1.133
$ sudo -E ./install_openvino_dependencies.sh## GUI version installer$ sudo ./install_GUI.sh
 or## CUI version installer$ sudo ./install.sh

2.Configure the Model Optimizer. Execute the following command.

$ cd /opt/intel/openvino/install_dependencies/
$ sudo -E ./install_openvino_dependencies.sh
$ nano ~/.bashrcsource /opt/intel/openvino/bin/setupvars.sh

$ source ~/.bashrc
$ cd /opt/intel/openvino/deployment_tools/model_optimizer/install_prerequisites/
$ sudo ./install_prerequisites.sh

3.【Optional execution】 Additional installation steps for the Intel® Movidius™ Neural Compute Stick v1 and Intel® Neural Compute Stick v2

$ sudo usermod -a -G users "$(whoami)"$ cat <<EOF > 97-usbboot.rulesSUBSYSTEM=="usb", ATTRS{idProduct}=="2150", ATTRS{idVendor}=="03e7", GROUP="users", MODE="0666", ENV{ID_MM_DEVICE_IGNORE}="1"SUBSYSTEM=="usb", ATTRS{idProduct}=="2485", ATTRS{idVendor}=="03e7", GROUP="users", MODE="0666", ENV{ID_MM_DEVICE_IGNORE}="1"SUBSYSTEM=="usb", ATTRS{idProduct}=="f63b", ATTRS{idVendor}=="03e7", GROUP="users", MODE="0666", ENV{ID_MM_DEVICE_IGNORE}="1"EOF$ sudo cp 97-usbboot.rules /etc/udev/rules.d/
$ sudo udevadm control --reload-rules
$ sudo udevadm trigger
$ sudo ldconfig
$ rm 97-usbboot.rules

4.【Optional execution】 Additional installation steps for processor graphics (GPU, Intel HD Graphics series only)

$ cd /opt/intel/openvino/install_dependencies/
$ sudo -E su
$ uname -r
4.15.0-42-generic #<--- display kernel version sample### Execute only when the kernel version is older than 4.14$ ./install_4_14_kernel.sh

$ ./install_NEO_OCL_driver.sh
$ sudo reboot

2. Work with RaspberryPi (Raspbian Stretch)

[Note] Only the execution environment is introduced.

1.Execute the following command.

$ sudo apt update
$ sudo apt upgrade
$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1NFt6g6ZkneHioU2P7rUJ8BFpQhIazbym" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1NFt6g6ZkneHioU2P7rUJ8BFpQhIazbym" -o l_openvino_toolkit_raspbi_p_2019.1.133.tgz
$ tar -zxvf l_openvino_toolkit_raspbi_p_2019.1.133.tgz
$ rm l_openvino_toolkit_raspbi_p_2019.1.133.tgz
$ sed -i "s|<INSTALLDIR>|$(pwd)/inference_engine_vpu_arm|" inference_engine_vpu_arm/bin/setupvars.sh

2.Execute the following command.

$ nano ~/.bashrc### Add 1 row belowsource /home/pi/inference_engine_vpu_arm/bin/setupvars.sh

$ source ~/.bashrc### Successful if displayed as below[setupvars.sh] OpenVINO environment initialized

$ sudo usermod -a -G users "$(whoami)"$ sudo reboot

3.Update USB rule.

$ sh inference_engine_vpu_arm/install_dependencies/install_NCS_udev_rules.sh### It is displayed as followsUpdate udev rules so that the toolkit can communicate with your neural compute stick
[install_NCS_udev_rules.sh] udev rules installed

[Note] OpenCV 4.1.0 will be installed without permission when the work is finished. If you do not want to affect other environments, please edit environment variables after installation is completed.

Training with your own data set

See the article below.
A sample of one-class training with Darknet and tiny-YoloV3.
https://qiita.com/PINTO/items/7dd7135085a7249bf17a#support-for-local-training-and-openvino-of-one-class-tiny-yolov3-with-a-proprietary-data-set

How to install Bazel (version 0.17.2, x86_64 only)

1. Bazel introduction command

$ cd ~$ curl -sc /tmp/cookie "https://drive.google.com/uc?export=download&id=1dvR3pdM6vtkTWqeR-DpgVUoDV0EYWil5" > /dev/null
$ CODE="$(awk '/_warning_/ {print $NF}' /tmp/cookie)"$ curl -Lb /tmp/cookie "https://drive.google.com/uc?export=download&confirm=${CODE}&id=1dvR3pdM6vtkTWqeR-DpgVUoDV0EYWil5" -o bazel
$ sudo cp ./bazel /usr/local/bin
$ rm ./bazel

2. Supplementary information

https://github.com/PINTO0309/Bazel_bin.git

How to check the graph structure of a ".pb" file [Part.1]

Simple structure analysis.

1. Build and run graph structure analysis program

$ cd ~$ git clone -b v1.11.0 https://github.com/tensorflow/tensorflow.git
$ cd tensorflow
$ git checkout -b v1.11.0
$ bazel build tensorflow/tools/graph_transforms:summarize_graph
$ bazel-bin/tensorflow/tools/graph_transforms/summarize_graph --in_graph=xxxx.pb

2. Sample of display result

YoloV3

Found 1 possible inputs: (name=inputs, type=float(1), shape=[?,416,416,3]) 
No variables spotted.
Found 1 possible outputs: (name=output_boxes, op=ConcatV2) 
Found 62002034 (62.00M) const parameters, 0 (0) variable parameters, and 0 control_edges
Op types used: 536 Const, 372 Identity, 87 Mul, 75 Conv2D, 72 FusedBatchNorm, 72 Maximum, 28 Add, 
24 Reshape, 14 ConcatV2, 9 Sigmoid, 6 Tile, 6 Range, 5 Pad, 4 SplitV, 3 Pack, 3 RealDiv, 3 Fill, 
3 Exp, 3 BiasAdd, 2 ResizeNearestNeighbor, 2 Sub, 1 Placeholder
To use with tensorflow/tools/benchmark:benchmark_model try these arguments:
bazel run tensorflow/tools/benchmark:benchmark_model -- 
--graph=/home/b920405/git/OpenVINO-YoloV3/pbmodels/frozen_yolo_v3.pb 
--show_flops 
--input_layer=inputs 
--input_layer_type=float 
--input_layer_shape=-1,416,416,3 
--output_layer=output_boxes

tiny-YoloV3

Found 1 possible inputs: (name=inputs, type=float(1), shape=[?,416,416,3]) 
No variables spotted.
Found 1 possible outputs: (name=output_boxes, op=ConcatV2) 
Found 8858858 (8.86M) const parameters, 0 (0) variable parameters, and 0 control_edges
Op types used: 134 Const, 63 Identity, 21 Mul, 16 Reshape, 13 Conv2D, 11 FusedBatchNorm, 11 Maximum, 
10 ConcatV2, 6 Sigmoid, 6 MaxPool, 4 Tile, 4 Add, 4 Range, 3 RealDiv, 3 SplitV, 2 Pack, 2 Fill, 
2 Exp, 2 Sub, 2 BiasAdd, 1 Placeholder, 1 ResizeNearestNeighbor
To use with tensorflow/tools/benchmark:benchmark_model try these arguments:
bazel run tensorflow/tools/benchmark:benchmark_model -- 
--graph=/home/b920405/git/OpenVINO-YoloV3/pbmodels/frozen_tiny_yolo_v3.pb 
--show_flops 
--input_layer=inputs 
--input_layer_type=float 
--input_layer_shape=-1,416,416,3 
--output_layer=output_boxes

How to check the graph structure of a ".pb" file [Part.2]

Convert to text format.

1. Run graph structure analysis program

$ python3 tfconverter.py### ".pbtxt" in ProtocolBuffer format is output.### The size of the generated text file is huge.

How to check the graph structure of a ".pb" file [Part.3]

Use Tensorboard.

1. Run log output program for Tensorboard

import tensorflow as tffrom tensorflow.python.platform import gfilewith tf.Session() as sess:
    model_filename ="xxxx.pb"
    with gfile.FastGFile(model_filename, "rb") as f:
        graph_def = tf.GraphDef()
        graph_def.ParseFromString(f.read())
        g_in = tf.import_graph_def(graph_def)    LOGDIR="path/to/logs"
    train_writer = tf.summary.FileWriter(LOGDIR)
    train_writer.add_graph(sess.graph)

2. Starting Tensorboard

$ tensorboard --logdir=path/to/logs

3. Display of Tensorboard

Access http://localhost:6006 from the browser.

How to check the graph structure of a ".pb" file [Part.4]

Use netron.

1. Install netron

$ sudo -H pip3 install netron

2. Starting netron

$ netron -b [MODEL_FILE]

3. Display of netron

Access http://localhost:8080 from the browser.

Neural Compute Stick 2

https://ncsforum.movidius.com/discussion/1302/intel-neural-compute-stick-2-information

Issue

OpenVINO failing on YoloV3's YoloRegion, only one working on FP16, all working on FP32
Regarding YOLO family networks on NCS2. Possibly a work-around
Convert YOLOv3 Model to IR

Reference

https://github.com/opencv/opencv/wiki/Intel%27s-Deep-Learning-Inference-Engine-backendhttps://github.com/opencv/opencv/wiki/Intel%27s-Deep-Learning-Inference-Engine-backend#raspbian-stretch