Thanks! It was a tough one too

GenPT's generative capabilities can be leveraged to improve point trajectory estimates by utilizing a best-first search on generated samples during inference, guided by the model’s own confidence of its predictions.

GenPT is trained with a modified flow matching setup, incorporating iterative refinements, a window-dependent prior for cross-window consistency, and a variance schedule tuned for point coordinates.

It is a generative framework for modelling multi-modal point trajectories.

Tracking the location of a point through a video is an ill-posed problem, where multiple plausible solutions can arise during visual obfuscations, such as occlusion.

Current point trackers are discriminative and fail to model the multi-modality inherent in point tracking.

Uhhh wat

I think getting into any PhD for lucrativeness hasn’t really ever been a good idea. Personally, the challenge I’ve been facing with my AI PhD has been how industry competition has negatively affected the ability to publish. Resource requirements are at an all time high for most types of projects.

I wouldn’t pivot just for pivot’s sake though. If you feel that you’re enjoying the direction you’re taking, then stick with it. That being said, if you feel stuck in the mud for a couple of years, then it might be best to pivot.

I feel seen. I’m going on to my sixth year and it’s been an immense struggle. I’ve missed many opportunities due to not working on LLMs, and many more due to working on self-organizing models. So I decided to pivot to a more traditional vision subject: point tracking. It’s been a demoralizing PhD…

I’m going to guess about 2 TB worth of keynote files.