Very cool, couple of additional ideas

[lonnylundsten]

This is great! I've been looking for something like this for quite a while. I have two suggestions:

1. Output results in yolo txt format similar to RectLabel:
   https://rectlabel.com

2. Allow for inference from a live video feed:
   https://developer.apple.com/documentation/vision/recognizing_objects_in_live_capture
   a. webcam for testing purposes
   b. video capture card BMD decklink and ultrastudio: https://www.blackmagicdesign.com/products

3. Allow user to select how often inference is run on video, i.e., perhaps instead of running on all frames a user could select 1 frame per second, etc.

[npna]

Hi, thanks for the suggestions.

I'll try to improve the app during the weekends with these in mind.