Self-Supervised Contrastive Learning and GAN-Based Denoising for High-Fidelity HumanNeRF Images
Qian Xu, Wenxuan Xu, Meng Huang, Weiqing Yan, Yang Guo · 2025
Read Self-Supervised Contrastive Learning and GAN-Based Denoising for High-Fidelity HumanNeRF Images by Qian Xu, Wenxuan Xu, Meng Huang, Weiqing Yan, Yang Guo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-12788205.txt.
Description
Qian Xu and four co-authors clean up the noisy renders produced by HumanNeRF, a system that rebuilds moving people in three dimensions. Its images suffer noise and lost detail from incomplete training data and rendering sampling. Their denoiser needs no labeled clean images.
Self-supervised contrastive learning first teaches the network to tell noise patterns from genuine human detail by comparing constructed sample pairs. A generative adversarial network then sharpens the result, its discriminator pushing the generator toward realism. Experiments showed effective noise removal with significantly better detail fidelity. Cleaner renders feed directly into more convincing three-dimensional human reconstruction for film, games, and telepresence.
Catalog record
| CREATOR | Qian Xu, Wenxuan Xu, Meng Huang, Weiqing Yan, Yang Guo |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |