I think this community is more LLM focused than computer vision, but I’m hoping it’s ok to post this here.

I struggled my way through getting tensorflow setup, and getting a model trained - it took about 10 hours over a few days, cross referencing different articles and videos, fighting to get protobufs compiled, and images/annotations converted to TFRecords. I finally got a basic model, but it was a nightmare, and I’m not sure I could figure it out again if I needed to.

Then I stumbled on this guy’s yolov8 object detection video. It was so easy. I had a trained model in less than an hour. I would highly recommend.

Also worth noting - the ultralytics folks have been very helpful on their discord server.

I’m not affiliated with the guy making the videos or the ultralytics team, I just wanted to plug them since they’ve been very helpful to me.

If you want you dip your feet in, and you have any basic questions, feel free to ask them here. I’ll answer any that i can.

Edit:

A quick note: In the video he uses an online tool for labeling - it looks like it can be installed locally, but it looks like a fair bit of work. I use label-studio which can be easily installed with pip.