License-Plate-Detection

Automatic License Plate Detection using YOLOv4 and Darknet

Files

Bounding_Box_Visualization.ipynb contains the notebook to view the analysis of bounding-box prediction model.

Plate_Detection_YOLOV4_Darknet_.ipynb contains the notebook to view the best-evaluated Yolov4 model.

Approach

• Cities with lakhs of vehicles running on the roads cannot afford the inadequate manual method of license plate detection.
• Moving objects can be detected using adaptive background subtraction.
• Edge detection algorithm is used to get segmented moving objects.
• YOLOv2, YOLOv3, YOLOv4 or COCO dataset can be employed to detect different types of objects.
• For detecting license plates and extracting the characters several methods have been tested and evaluated, such as OCR, MobileNets and Inception-v3, Open ALPR.

Result

License Plate Detection Using OCR and Pytesseract Library

Helmet detection using YoloV3 model(Pytorch):

Automatic Detection of Bike Rider without Helmet using TensorFlow API:

Conclusion

• A centroid tracking method was also proposed to reduce the number of false positives generated by the helmeted bikers when their helmet is out of video frame.
• Chances of death due to road accident will reduce.
• Real-world video surveillance system that can effectively detect moving person using limited resources.