Convert Scaled YoloV4

[at1993]

Hi TNTWEN,

As an enhancement request. Will you be supporting Scaled Yolo V4? This will be nice thanks.

[TNTWEN]

@at1993

I have finished the adaptation of yolov4-csp，But the usage is very different from the previous one. I'll publish it in a new branch later.
And more models are also comming

[TNTWEN]

@at1993
https://github.com/TNTWEN/OpenVINO-YOLOV4/tree/ScaledYOLOv4

yolov4-csp AND yolov4x-mish have been uploaded

When darknet support yolov4-P5,yolov4-P6,yolov4-P7 ,i will also update them

[at1993]

Awesome! Thanks.

[at1993]

Hi TNTWEN, I got the following error when running the script

python "C:\Program Files (x86)\Intel\openvino_2021.3.394\deployment_tools\model_optimizer\mo.py" --input_model frozen_darknet_yolov4_model.pb --batch 1 --reverse_input_channels --output detector/yolo-v4/Conv_94/BiasAdd,detector/yolo-v4/Conv_104/BiasAdd,detector/yolo-v4/Conv_114/BiasAdd/

D:\Programming\Python\OpenVINO-YOLOV4-Scaled\yolov4-csp>python "C:\Program Files (x86)\Intel\openvino_2021.3.394\deployment_tools\model_optimizer\mo.py" --input_model frozen_darknet_yolov4_model.pb --batch 1 --reverse_input_channels --output detector/yolo-v4/Conv_94/BiasAdd,detector/yolo-v4/Conv_104/BiasAdd,detector/yolo-v4/Conv_114/BiasAdd/
Model Optimizer arguments:
Common parameters:
- Path to the Input Model: D:\Programming\Python\OpenVINO-YOLOV4-Scaled\yolov4-csp\frozen_darknet_yolov4_model.pb
- Path for generated IR: D:\Programming\Python\OpenVINO-YOLOV4-Scaled\yolov4-csp.
- IR output name: frozen_darknet_yolov4_model
- Log level: ERROR
- Batch: 1
- Input layers: Not specified, inherited from the model
- Output layers: detector/yolo-v4/Conv_94/BiasAdd,detector/yolo-v4/Conv_104/BiasAdd,detector/yolo-v4/Conv_114/BiasAdd/
- Input shapes: Not specified, inherited from the model
- Mean values: Not specified
- Scale values: Not specified
- Scale factor: Not specified
- Precision of IR: FP32
- Enable fusing: True
- Enable grouped convolutions fusing: True
- Move mean values to preprocess section: None
- Reverse input channels: True
TensorFlow specific parameters:
- Input model in text protobuf format: False
- Path to model dump for TensorBoard: None
- List of shared libraries with TensorFlow custom layers implementation: None
- Update the configuration file with input/output node names: None
- Use configuration file used to generate the model with Object Detection API: None
- Use the config file: None
- Inference Engine found in: C:\Program Files (x86)\Intel\openvino_2021\python\python3.6\openvino
Inference Engine version: 2.1.2021.3.0-2787-60059f2c755-releases/2021/3
Model Optimizer version: 2021.3.0-2787-60059f2c755-releases/2021/3
[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.user_data_repack.UserDataRepack'>): No node with name detector/yolo-v4/Conv_114/BiasAdd/

[TNTWEN]

python "C:\Program Files (x86)\Intel\openvino_2021.3.394\deployment_tools\model_optimizer\mo.py" --input_model frozen_darknet_yolov4_model.pb --batch 1 --reverse_input_channels --output detector/yolo-v4/Conv_94/BiasAdd,detector/yolo-v4/Conv_104/BiasAdd,detector/yolo-v4/Conv_114/BiasAdd/

you should delete the last /

[TNTWEN]

python "C:\Program Files (x86)\Intel\openvino_2021.3.394\deployment_tools\model_optimizer\mo.py" --input_model frozen_darknet_yolov4_model.pb --batch 1 --reverse_input_channels --output detector/yolo-v4/Conv_94/BiasAdd,detector/yolo-v4/Conv_104/BiasAdd,detector/yolo-v4/Conv_114/BiasAdd

This is correct

[at1993]

Thanks, it works now. The python program is running OK, However, when I run it with the C++ object_detection_demo program, it returns an error.

\Documents\Intel\OpenVINO\omz_demos_build\intel64\Release>object_detection_demo -m "D:\Programming\Python\OpenVINO-YOLOV4-Scaled\yolov4-csp\frozen_darknet_yolov4_model.xml" -i "waikikibeach.mp4" -d GPU -at yolo
[ INFO ] InferenceEngine: API version ......... 2.1
Build ........... 2021.3.0-2787-60059f2c755-releases/2021/3
[ INFO ] Parsing input parameters
[ INFO ] Reading input
MFX: Can't initialize session
[ INFO ] Loading Inference Engine
[ INFO ] Device info:
[ INFO ] GPU
clDNNPlugin version ......... 2.1
Build ........... 2021.3.0-2787-60059f2c755-releases/2021/3
Loading network files
[ INFO ] Batch size is forced to 1.
[ INFO ] Checking that the inputs are as the demo expects
[ INFO ] Checking that the outputs are as the demo expects
[ ERROR ] Invalid output type: Add. RegionYolo expected

[TNTWEN]

This is the difference between Scaled-yolov4 and yolov4 . Model conversion and inference demo need special treatment.
So it is not compatible with openvino's official C + + demo

[at1993]

Do you have any sample code in C++ for this "special treatment" as my library is written in C++? Would love to be able to run it in C++.

[TNTWEN]

I'm not familiar with C++ demo，but i will try

[at1993]

Thank you.

[RohitKeshari]

Any one find any solution for openvino version of Scaled-yolov4?

[TNTWEN]

@RohitKeshari This repo already supports Scaled-yolov4
https://github.com/TNTWEN/OpenVINO-YOLOV4/tree/ScaledYOLOv4