George Cazenavette (@gcazenavette) 's Twitter Profile
George Cazenavette

@gcazenavette

Vision stuff at @MIT_CSAIL

ID: 1260477645759942656

calendar_today13-05-2020 07:51:13

96 Tweet

240 Followers

255 Following

Vincent Sitzmann (@vincesitzmann) 's Twitter Profile Photo

How can we learn to generate 3D scenes directly with diffusion models if we only have images, no ground-truth 3d scenes? Ayush, Tianwei and George will tell you at our poster “diffusion with Forward Models”, #202!

How can we learn to generate 3D scenes directly with diffusion models if we only have images, no ground-truth 3d scenes? Ayush, Tianwei and George will tell you at our poster “diffusion with Forward Models”, #202!
Vincent Sitzmann (@vincesitzmann) 's Twitter Profile Photo

Introducing pixelSplat: feed-forward Gaussian splats from image pairs! Led by David Charatan and Lester Li, collaborating with Andrea Tagliasacchi 🇨🇦! We propose a memory-efficient, fast and editable alternative to pixelNeRF based on 3D Gaussian Splatting! davidcharatan.com/pixelsplat/ 1/n

Andrea Tagliasacchi 🇨🇦 (@taiyasaki) 's Twitter Profile Photo

📢📢📢 introducing 𝐩𝐢𝐱𝐞𝐥𝐒𝐩𝐥𝐚𝐭: feed-forward Gaussian splats from image pairs! pixelNeRF + 3D Gaussian Splatting == davidcharatan.com/pixelsplat Led by David Charatan and S. Lester Li, collaboration with Vincent Sitzmann

George Cazenavette (@gcazenavette) 's Twitter Profile Photo

.Tong’s ReparamModule is also how we made differentiable updates work with a learnable step size in our Dataset Distillation paper. It essentially allows you to take an existing pytorch module f_\theta(x) and reparameterize it as f(x, \theta). Very useful!

Vincent Sitzmann (@vincesitzmann) 's Twitter Profile Photo

Introducing “FlowMap”, the first self-supervised, differentiable structure-from-motion method that is competitive with conventional SfM like Colmap! cameronosmith.github.io/flowmap/ IMO this solves a major missing piece for internet-scale training of 3D Deep Learning methods. 1/n