Using YOLO9000, classify objects in a video (made to support a wider range of computers than YOLO-9000 supports)
- Python >= 3.5
- Yolo-9000
- FFMPEG
- Run
ffmpeg -i video.mp4 "frames/out-%08d.jpg"
on your desired video - Put the frames folder into YOLO-9000's folder
- Put
strain.py
intodarknet/
and run the script - Concatenate the frames back into a video with
ffmpeg -framerate 25 -i out_%08d.jpg output.mp4
(framerate will need adjusting)
- The script takes a long time to run. This is because the nueral network is running on the CPU for compatability reasons. Machine learning is much faster on a GPU but very few GPUs fully support NN/ML. To remedy, run overnight.
I collected clips of Sydney and fed it through the classifier and paired it with music to produce this result.