Reproduce “ResPCC: A Loss-Resilient Neural Point Cloud Codec over Lossy Networks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xueqin Niu, Mufan Liu, Yifan Wang, Le Yang, Jun Sun, Yiling Xu
ResPCC, a neural point cloud codec developed by Shanghai Jiaotong University and the University of Canterbury, incorporates intrinsic resilience against packet loss in its end-to-end design. It consistently delivers higher reconstruction quality at comparable or lower bitrates across varying packet loss rates, outperforming established traditional and learning-based methods by up to 19.31% in D2-PSNR on ShapeNet.
Key insights
- Designing packet-loss-aware mechanisms directly into the neural compression pipeline enables intrinsic resilience, moving beyond external protection or reactive concealment strategies.
- Dynamically modulating latent feature generation at the encoder based on predicted loss rates allows for adaptive performance across diverse network conditions with a single model.
- Transforming channel-wise data loss into spatially scattered element-wise missing patterns via spatial-channel interleaving simplifies the subsequent restoration task for the decoder.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.