Using AutoML to train the object localization task.
To train the model, we should prepare the dataset. The file capture.py can be used to record the video using cv2.VideoCapture()
. After getting the recording video, feel free to use ffmpeg
to extract the frames Noted that ffmpeg will overwrite the existing images. Alternativly, readers can use extract.py to extract the image without overwritting the images.
# Using ffmpeg
ffmpeg -i output.mp4 -vf "fps=20" "frame/out%02d.jpg"
# Using extract.py
python extract.py -n <input file> -l <location of output files> -p <prefix of extracted files>