Summary of Multi-hypotheses Conditioned Point Cloud Diffusion For 3d Human Reconstruction From Occluded Images, by Donghwan Kim et al.
Multi-hypotheses Conditioned Point Cloud Diffusion for 3D Human Reconstruction from Occluded Images
by Donghwan Kim, Tae-Kyun Kim
First submitted to arxiv on: 27 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel pipeline, MHCDIFF, for 3D human shape reconstruction under severe occlusion. Traditional parametric models like SMPL(-X) are limited to minimally-clothed human shapes and struggle with misaligned inputs. Implicit-function-based methods can capture geometric details but fail to handle occluded regions given a single RGB image. MHCDIFF combines point cloud diffusion conditioned on probabilistic distributions for pixel-aligned detailed 3D reconstruction. The method extracts local features from multiple hypothesized SMPL(-X) meshes, aggregates the set of features, and conditions the diffusion model. In experiments on CAPE and MultiHuman datasets, MHCDIFF outperforms various state-of-the-art methods under synthetic and real occlusions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research is about creating better 3D models of people from pictures. When there’s a lot of clothing or hair in the way, it can be hard to get an accurate model. The authors propose a new method that can handle these situations by combining different parts of the picture and using probability distributions to make sure everything lines up. They tested this method on two large datasets and found that it works better than other methods for creating 3D models of people. |
Keywords
» Artificial intelligence » Diffusion » Diffusion model » Probability
