ANPR_IR

Iranian Cars Number Plate

STEP_1: Plate Detection with YOLOV7

Run train_Yolov7.ipynb

Dataset01: https://github.com/roozbehrajabi/ALPR_Dataset /(YOLOv3_Dataset)

Dataset02: https://www.kaggle.com/datasets/skhalili/iraniancarnumberplate

Steps to download Kaggle datasets in Google Colab: https://www.kaggle.com/general/156610

Result train_Yolov7.ipynb:

P = 0.985 -- R = 0.992 -- [email protected] = 0.995 -- [email protected]:.95 = 0.779

Detect Plate:

Crop Plate Region:

STEP_2: OCR Plate Numbers with fasterrcnn_resnet50

Run tarin_ocr_fasterrcnn.ipynb

Dataset01: https://github.com/roozbehrajabi/ALPR_Dataset /(Faster_R-CNN_dataset)

Install Detecto package for train fasterrcnn_resnet50:

https://detecto.readthedocs.io/en/latest/index.html

pip install detecto

OCR Plate Numbers:

ANPR: Yolov7_ANPR.py

Cloning Yolov7 repository in "/ANPR_IR/ANPR/" directory

Yolov7: https://github.com/WongKinYiu/yolov7

pip install deep-sort-realtime

pip install detecto

change font path in "/yolov7-main/utils/plots.py/def plot_one_box_PIL()" like "/ANPR_IR/ANPR/plots.py"

Run Yolov7_ANPR.py

Test Image ANPR: get_plates_from_image(plate_image)

Test Video ANPR: get_plates_from_video(video_path)

Test Webcam ANPR: get_plates_from_webcam()

ANPR: