Using the training images located in CarPartsDetectionChallenge/Data/Source_Images/Training_Images and the annotation file data_train.txt which we have created in the previous step we are now ready to train our YOLOv3 detector.
Before getting started download the pre-trained YOLOv3 weights and convert them to the keras format. To run both steps run the download and conversion script from within the CarPartsDetectionChallenge/2_Training directory:
python Download_and_Convert_YOLO_weights.py
To list available command line options run python Download_and_Convert_YOLO_weights.py -h.
To start the training, run the training script from within the CarPartsDetectionChallenge/2_Training directory:
python Train_YOLO.py
Depending on your set-up, this process can take a few minutes to a few hours. The final weights are saved in CarPartsDetectionChallenge/Data/Model_weights. To list available command line options run python Train_YOLO.py -h.
If training is too slow on your local machine, consider using cloud computing services such as AWS to speed things up. To learn more about training on AWS navigate to CarPartsDetectionChallenge/2_Training/AWS.
Next, go to CarPartsDetectionChallenge/3_Inference to test your YOLO detector on new images!