The implemetation uses YoloV3 from darknet framework, pytorch, tensorflow-gpu and keras
-
Install darknet:
cd darknet && make clean
make all
cp -f libdarknet.so ../.
cd ..
-
Create an anaconda environment, using
conda create -n <env_name>
. If you don't have anaconda installed, install anaconda -
Activate anaconda environment using
conda activate <env_name>
-
Run
source install.txt
will install all the necessary conda packages. Just pressy
whenever prompted
python main.py /path/to/input/video/file
Output will be saved in output/<input_name>_result_tracker<input_img_size>.avi
. The detected number plates will be stored at output/license_plates
. The detected cars will be stored at output/cars
. Average FPS will be printed on the console after output has been saved.
python main.py -h
for more options