rPPG-VQA: A Video Quality Assessment Framework for Unsupervised rPPG Training

rPPG-VQA addresses a specific technical bottleneck in the remote biometric sensing niche: the 'garbage in, garbage out' problem of unsupervised training on noisy video. While traditional Video Quality Assessment (VQA) focuses on human visual perception (e.g., blur, compression), this project focuses on signal integrity for sub-perceptual skin color changes. With 0 stars and 4 forks at age 4 days, this is currently a raw academic reference implementation (arXiv:2604.11156). Its defensibility is very low because the value lies in the algorithm/methodology described in the paper, which can be easily replicated by researchers in the field. There is no community, data moat, or significant software engineering complexity beyond standard PyTorch vision pipelines. Frontier labs like OpenAI or Google are unlikely to build this directly as it is too domain-specific for general-purpose foundation models. However, specialized health-tech players like Binah.ai or research groups working on PhysBench could absorb this logic. The displacement horizon is relatively short (1-2 years) as unsupervised rPPG is a fast-moving research area where newer, noise-invariant architectures might eventually make pre-training quality filtering obsolete.