license-plate-detector
Using feature extraction and CNNs, this project reads in images of cars, locates the license plate on the car, and reads the license plate number.
Our report can be found in docs/final_report.pdf.
To run:
-
Navigate to the code directory
-
Run the command python3 main.py --generate-weights which will train the model from scratch and save model weights to the saved_weights folder.
-
Test model using the --load-weight and --test-uploaded-image flags. A few examples that can be run are:
python3 main.py --load-weight <path_to_weight_in_saved_weights_dir> --test-uploaded-image data_license_only/crop_h1/I00016.png ^expected output: 4B27433
python3 main.py --load-weight <path_to_weight_in_saved_weights_dir> --test-uploaded-image data_license_only/crop_h2/I00055.png ^expected output: 1BA0888
python3 main.py --load-weight <path_to_weight_in_saved_weights_dir> --test-uploaded-image data_license_only/crop_h4/I00011.png ^expected output: 3M46918
-
If --test-uploaded-image is not just an image of a license plate and requires plate detection to be run as well, be sure to specify the --needs-detector flag as well!