Introducing Slice3D (#CVPR2024 paper #23, 10:30-12, Arch 4A-E) for single-view 3D reconstruction by first predicting multi-slice images. The key is occlusion revelation, while sidestepping multi-view inconsistency. Trained on single A40 & inf time < 20s. yizhiwang96.github.io/Slice3D/

Thank you Zhengzhong for sharing! Our code is now avilable at github.com/mingrui-zhao/S…

Introducing SMITE: A diffusion-based network segmenting an object in a video with any given granularity without losing temporal consistency thanks to a voting mechanism established by tracking over the image and influencing the attention maps. Project: segment-me-in-time.github.io