Have a random thought to use opencv to filter video clips.
The idea is to first use ffmpeg to break recordings into 2 minutes chunk, then apply yolo to see if there is a bird or some other interesting object shows up. If not, delete the clip.
Then find out the yolo model is not maintained by one people. And the javacv is not that intuitive to use. Might be too hig to fit in a weekend, but definitely gonna have a try.