discussion / Camera Traps  / 25 March 2024

Integrating AI models with camera trap management applications

Hi All, As part of extending the work we are doing at the BearID Project, we are thinking about integrating the models we are developing into open source camera trap project. This could include MegaDetector, swappable specias detectors and swappable individual reID. I'd like to aim for something that could be run locally or as a service, and possibly across platforms. I'd like to support images and videos. My leading candidates at the moment are TRAPPER and Camelot.

Any thoughts? Is someone already doing this?

Hey Ed! I'm not clear about what you're proposing, could you elaborate? Might be that a word is missing from your first sentence - like are you proposing to make your model accessible or are you proposing to integrate existing open models into bear id? Or are you building a system that allows things to swap in and out? Tell me more?  

My StalkedByTheState project runs AI inferences models as a service. I'm working on a client program that will be dockerised to use as a client. Currently I'm only doing this for yolo-like inference engines.

so you can do something like docker run my-engine /tmp/mydir my-service-url and it will classify all the images in there sort of thing.