As #ICCV2023 is just around the corner, I'd like to share our work. "ENVIDR: Implicit Differentiable Renderer with Neural Environment Lighting." The idea is to pre-train a color-rendering MLP used by neural fields to let it learn the rule of physically-based rendering. (1/n)

Really cool that you can transfer the lighting from one scene to another object! Check out Ruofan Liang work at ICCV! Proud to share the same group under Nandita Vijaykumar :)

Thrilled to share our ICCV work -- kudos to my students Ruofan Liang Fan Chen and our collaborators at Intel!

Another proud advisor moment! My student Sankeerth Sankeerth Durvasula presenting his work on accelerating CNN-based perception for event cameras at IROS 2023! Leverages this neat idea of only operating on differences between inputs. #IROS2023

We (cc Caroline Trippel) congratulate the graduate and undergraduate finalists in Association for Computing Machinery Student Research Competition at MICRO 2023! Thank you to all participants for submitting excellent work and to the Selection and Judging Committees for your service.

This is a paper I am pretty excited about as I was working some time on speeding up #3DGS training. github.com/MrNeRF/gaussia… "DISTWAR: Fast Differentiable Rendering on Raster-based Rendering Pipelines" Paper: arxiv.org/pdf/2401.05345…

3D Gaussian offers fast and accurate scene representations, but it is still inefficient due to atomic operations. Our work, DISTWAR, makes 3D Gaussian Splatting even faster - ~2.3X on RTX 4090. Our paper and code that works out-of-the-box: github.com/Accelsnow/gaus… #3DGS