Reproduce “Efficient Remote Sensing Instance Segmentation with Linear-Time State Space Distilled Visual Foundation Models” 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
Qinzhe Yang, Keyan Chen, Jia Xu, Zhenwei Shi, Zhengxia Zou
RS4D transfers SAM’s visual features into a smaller state-space encoder through feature distillation, then fine-tunes it for remote-sensing instance segmentation. Its linear-scaling sequence mixing avoids Transformer self-attention’s quadratic token-count cost. On SSDD, it reports 70.7 box AP and 70.0 mask AP, with 11.75M backbone parameters and 108.8G FLOPs versus ViT-Base’s 89.67M and 972G. Results are weaker on NWPU and do not establish deployment speedups.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.