FurE speeds up realistic 3D animal fur reconstruction by ten times
FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets
Computer Vision and Pattern RecognitionArtificial IntelligenceGraphics
Summary
Creating detailed 3D models of animal fur from pictures is hard because of tiny hairs, overlapping, and no big sets of animal hair pictures to learn from. The authors developed FurE, a way to rebuild each hair strand efficiently by using a kind of model trained on human hair data instead of animal hair data. FurE can quickly create editable, detailed fur that matches the real animal without needing a lot of training time. This method works fast and well, even on both computer-generated and real animal images.
What this means in practice
- •For visual effects artists: Generate detailed and editable 3D animal fur models rapidly for movies and animation from multiple images without needing specialized animal fur datasets.$Commercial implications: Enables studios to create realistic fur assets faster and less expensively for visual effects in films or games.
- •For wildlife documentary producers: Produce highly realistic, customizable 3D animal fur reconstructions from video footage for immersive educational content and visual storytelling.
Authors
Srinjay Sarkar, Prakhar Kaushik, Soumava Paul, Alan Yuille
Abstract
Realistic 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. We present FurE, an efficient strand-based animal fur reconstruction method that recovers a per-strand, editable groom by optimizing a root-conditioned latent field, decoded into strand geometry via a PCA-based decoder. We reconstruct a defurred animal body using local fur-thickness cues from a surface-constrained Gaussian Frosting representation together with part-based priors. We further show that a PCA-based decoder learned from human-hair strand data can alleviate animal-data scarcity while enabling substantially faster optimization. FurE achieves a 10x speedup in strand training over current SOTA dense per-strand optimization while retaining strand fidelity and generalizing across synthetic and real-world sequences, with quantitative and qualitative validation despite the reduction in training time.