Zander Lewis
Check their projects out: Lectern, Wasteland, Hyperlight
Once you ship this you can't edit the description of the project, but you'll be able to add more devlogs and re-ship it as you add new features!
First devlog yay :)
Alright, so a lot of this time was spent on labeling and training the model. There's ~65 images (45 train, 20 val). I trained the model for 100 epochs. As of right now, there is a really nice mAP50 and good precision/recall. However, the model is really lacking in the mAP50-95.
This is my first actual attempt at a computer vision project, so I still don't understand what much of this means. Anyway, here's a video of the results that I have gotten thus far.
(I am aiming to get a lot more images for more training data later)