HKU open-sources GenSfM: jointly reconstructing 3D objects and camera poses from four uncalibrated images
A team at the University of Hong Kong has open-sourced GenSfM, which uses a native 3D diffusion model to jointly denoise object shape and per-view camera poses within a single multimodal sequence, requiring only 2 to 4 uncalibrated images and no camera intrinsics or extrinsics. The paper evaluates on Toys4k and GSO; compared with the multi-view baseline TRELLIS-M, novel-view PSNR improves by 3.2–4.5 dB, median Chamfer distance decreases by about 64%–74%, and pose Acc@5° exceeds 96%.
Highlights
HKU open-sources GenSfM: jointly reconstructing 3D objects and camera poses from four uncalibrated images
A team at the University of Hong Kong has open-sourced GenSfM, which uses a native 3D diffusion model to jointly denoise object shape and per-view camera poses within a single multimodal sequence, requiring only 2 to 4 uncalibrated images and no camera intrinsics or extrinsics. The paper evaluates on Toys4k and GSO; compared with the multi-view baseline TRELLIS-M, novel-view PSNR improves by 3.2–4.5 dB, median Chamfer distance decreases by about 64%–74%, and pose Acc@5° exceeds 96%.
3D视觉工坊 · 微信公众号Source ↗
Paper research
HKU open-sources GenSfM: jointly reconstructing 3D objects and camera poses from four uncalibrated images
A team at the University of Hong Kong has open-sourced GenSfM, which uses a native 3D diffusion model to jointly denoise object shape and per-view camera poses within a single multimodal sequence, requiring only 2 to 4 uncalibrated images and no camera intrinsics or extrinsics. The paper evaluates on Toys4k and GSO; compared with the multi-view baseline TRELLIS-M, novel-view PSNR improves by 3.2–4.5 dB, median Chamfer distance decreases by about 64%–74%, and pose Acc@5° exceeds 96%.
3D视觉工坊 · 微信公众号Source ↗