Hi, I’m Mikhail Okunev, 4th year PhD candidate in Brown Visual Computing Lab advised by James Tompkin. I’m broadly interested in 3D reconstruction, inverse rendering and dynamic scene representations, especially with a monocular camera.

I joined academia pretty late in life. In the past I had a career as a research/machine learning engineer in Meta Reality Labs, Meta spam detection team, Microsoft Bing and a Silicon Valley startup Instrumental. I’ve been working on a broad range of topics including lighting estimation, foveated rendering, video superresolution, automatic visual anomaly detection, spam & fraud detection, ranking, and etc.

In my free time I enjoy brewing coffee, lifting weights, and playing piano.

Publications

F-TöRF: Flowed Time of Flight Radiance Fields Mikhail Okunev*, Marc Mapeke*, Benjamin Attal, Christian Richardt, Matthew O'Toole, James Tompkin tl;dr: C-ToF depth cameras can't reconstruct dynamic objects well. We fix that with our NeRF model that takes raw ToF signal and reconstructs motion along with the depth. All with a static monocular camera! ECCV 2024 [Paper][Project Page]

Monocular Dynamic Gaussian Splatting is Fast and Brittle but Smooth Motion Helps Yiqing Liang, Mikhail Okunev, Mikaela Angelina Uy, Runfeng Li, Leonidas J. Guibas, James Tompkin, Adam Harley tl;dr: We test various SOTA Gaussian Splatting methods for dynamic monocular reconstruction and look at their performance on popular benchmarks. In submission [Paper]

Spatiotemporally Consistent HDR Indoor Lighting Estimation Zhengqin Li, Li Yu, Mikhail Okunev, Manmohan Chandraker, Zhao Dong tl;dr: Spatially and temporally consistent HDR lighting from videos ACM ToG, Presented in SIGGRAPH Asia 2023 [Paper][Video][Project Page]

DeepFovea: Neural Reconstruction for Foveated Rendering and Video Compression using Learned Statistics of Natural Videos Anton Kaplanyan, Anton Sochenov, Thomas Leimkühler, Mikhail Okunev, Todd Goodall, Gizem Rufo tl;dr: We made a foveated rendering system that allows us to render only ~10% of the pixels, while inpainting the rest with a neural network. SIGGRAPH Asia 2019 [Paper][Project Page]