Semantic Segmentation

[Waidhoferj]

Ideally, users of the web app should be able to interact with actual objects in the world, not just modals. To make this happen, the argo editor could "trace" objects with invisible hit-boxes so the frontend can identify when a user has clicks on, say, a door to a new location.

Find a model that can be applied to our panoramic dataset to identify key objects around campus, like doors, booths, bulletin boards, etc. The model doesn't need to identify these things specifically, but ideally it can identify objects with similar characteristics.

Post models you find in the comments.