https://pip-tracker.netlify.app/
Partial copy of the client code used for PipTracker.
I've removed unnecessary experimental models I was using and also removed tracking code.
This offers a few pages:
- Landing page for interacting with images previously captured
- "Capture" view with live camera
- Edit/preview view which lets a user choose how they may want to change an images capture groups and whether to save or discard
- A benchmark view for testing performance of new models
- A "rules" view for users to see rules per game/amount of players
The code is very much incomplete but there were many things I learned in the process and I feel obligated to share with the community about how to use all of these things together.
The primary app uses a YOLOv5 model since that had the best performance/latency for use on mobile devices.
The export of the model is slightly different than their normal export. I noticed I was able to squeeze better performance out when I ran the NMS with supplied JS code in YoloExperimentalV5Model instead of using the one that was packaged in the default export, which is used in the YoloObjectDetectionModel.