FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets
FurE reconstructs editable animal fur strands from images without animal-fur datasets, about 10x faster.
FurE reconstructs editable, per-strand animal fur from multi-view images without animal-fur datasets. It optimizes a root-conditioned latent field decoded into strand geometry by a PCA decoder, and recovers a defurred body using fur-thickness cues from a surface-constrained Gaussian Frosting representation plus part-based priors. A decoder learned from human-hair strands reduces animal-data scarcity and speeds training. The authors report a 10x strand-training speedup over dense per-strand optimization while retaining fidelity on synthetic and real sequences.
- Optimizes a root-conditioned latent field into editable fur strands.
- A human-hair PCA decoder avoids scarce animal-fur datasets.
- Reports about a 10x speedup versus dense per-strand optimization.
- Tested on synthetic and real multi-view animal sequences.
From the original article
1 min read at arxiv.orgRealistic and editable animal fur reconstruction from multi-view images is challenging due to fine-scale detail, self-occlusion and obfuscation, and, unlike human hair, the lack of animal-fur datasets. Fur usually covers most of an animal's body, with large inter-species and intra-species variability.
By Srinjay Sarkar · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.