Pyinstaller-OpenVINO-Norfair-Object-Detection-Tracking using custom YOLOv4 License Plate Detection model

Guide to convert custom YOLOv4 model to OpenVINO (IR) format

Convert yolov4 model (.weights) into Tensorflow (.pb) format

Download yolov4 .weights, .config, and .names files

https://drive.google.com/file/d/1ffeOmcTy7yDPOXSgrr_lR2bNSIBJCZK3/view?usp=sharing

Download video for inference

https://drive.google.com/file/d/1xxcud0qlQ4xcEgrjNntLGmNlxncuTZRU/view?usp=sharing

Clone the repo below & cd into it

https://github.com/theAIGuysCode/tensorflow-yolov4-tflite.git

Create and activate YOLOv4 Virtual environment.

python -m venv tf-yolov4
.\tf-yolov4\Scripts\activate

Install requirements

pip install -r requirements.txt

Edit __C.YOLO.CLASSES in '.\core\config.py' to path to appropriate classes

Place downloaded yolov4 model in '.\data'

Convert yolov4 model to TF format

python save_model.py --weights .\data\custom.weights --output .\checkpoints\custom-416 --input_size 416 --model yolov4 

Convert Tensorflow (.pb) model into Openvino IR model (.xml) format and run inference

Clone this repo & cd into it

Create and activate Openvino Virtual environment.

python -m venv yolov4-openvino
.\yolov4-openvino\Scripts\activate

Install requirements

pip install -r requirements.txt

Extract github.rar into repo

Move all files from'.\checkpoints\custom-416' in to '.\openvino\public\yolo-v4-tf\yolo-v4.savedmodel'

Move downloaded custom.weights file to '\openvino\public\yolo-v4-tf'

Convert yolov4 model to Openvino(IR) format

omz_converter --name yolo-v4-tf --download_dir openvino --precisions FP32

Build Pyinstaller '.exe' file

Install Pyinstaller

pip install pyinstaller

Create and edit '.spec' file then build the executable file

pyinstaller --add-data "lp.mp4;." --add-data "openvino\public\yolo-v4-tf\FP32\yolo-v4-tf.xml;." --add-data "obj.names;." --add-data "C:\Users\sapen\Desktop\Workspace\Roc4Tech\openvino\Lib\site-packages\openvino\*;openvino" --onefile openvino_norfair_object_detection.py

NB: Replace ';' with ':' for linux machines. Refer to this link for commands to customize .spec file: 'https://pyinstaller.org/en/stable/usage.html#what-to-bundle-where-to-search'

If build was successful, 'build' and 'dist' folders will be created. Inside the 'dist' folder you'll find the executable(.exe) file.

Run '.exe' file

.\dist\openvino_norfair_object_detection