Someone ate the pizza midway through recording so we got cardboard?

This is funny... If you look at the most cited paper, the first 19 pages of citations are self-citations. Interestingly, even if the publication date is 2022, there are self-citations dating back to 2017 (back to the future).

I guess one can make a clique of self citations and scale with N^2.

The ones from LLMs? 😂

Does the P in elephant stand for primitives?

That being said, Claude is proposing you to decompose the scene in animal-inspired primitives which sounds kinda fun 😂 a giraffe here, a lion over there, a dog underneath.

We looked a bit into this for the cross-descriptor matching paper, but I don't think the experiments made it to the supplementary. We tried some kind of cross-detector repeatability at different px thresholds (for several values of top K keypoints selected). We didn't consider scales though...

Great initiative! Just wondering: were we supposed to receive an email with the decision (maybe not everyone will be approved to attend due to space / other constraints) after registration or is that coming later?

What do you mean by flat? How are you writing?

I have files with >>100K entries (in a single group) and seems to work alright on my end (but I didn't benchmark in detail).

I guess the profs / senior supervisors could attend in person if on appropriate visas for publishing purposes, but first issue would still stand.

I remember hearing something about students on single entry visas when I first asked about CVPR's locations, but I have no clue if that's true. No (big) confs in the US would mean no opportunities for them to present their work in person (or even publish if in person attendance is mandatory).

Wait, I thought this is how it finds you 🤔

How many robots were hurt in the making of this video? 😂 Jokes aside, very impressive!