"AssertionError: The number of dimensions should be XXX." is the error message that always makes me wanna smash my keyboard and shutdown the whole server. Why people just can't spend another second logging the current shape of that thing?

I won't try auto-coding tools unless they can automatically solve the correct 3D camera transformation between frameworks and datasets, which is the unforgiven sin of we computer vision hackers

All you need for consistent video editing is consistent diffusion features! Check out TokenFlow with the greatest michal geyer Omer Bar Tal OneViTaDay

Me: Can you draw a very normal image? ChatGPT: Here is a very normal image depicting a tranquil suburban street scene during the daytime. Me: Not bad, but can you go more normal than that? (cont.)

The PhD-intern recruiting in tech giants are so broken now, it is easier to find internships on Twitter. It seems like the scientists cannot (or don't want to) see the candidate pool. I applied to Google and Apple at the end of last year, and I haven't heard anything back LoL

Huh. Weird. arxiv.org/abs/2312.04709

After six years in generative model research, my early works have become so old that they no longer get citations when people re-invent the same things. Surely nothing lasts forever, but I didn't expect them to die before my PhD defense... LOL

When will Gaussian Splatters start to realize they need an if statement to check the memory usage before densification?

さユりに関するお知らせ

At Creative Vision, Snap Research, we’re excited to announce our 2025 call for interns. Our past interns have built some of the world’s best image, video, and 3D generators. We’ve introduced new personalization methods and developed the world’s fastest mobile foundational image