Reproduce “MaskMerge: Structure-aware Token Merging for Efficient Instance Segmentation” 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
Bavesh Balaji, Letian Jiang, Alexander Wong, David Clausi, Sirisha Rambhatla
MaskMerge reduces cross-instance token merging in instance segmentation by using coarse object masks to exclude candidate token pairs assigned to different objects. On PanNuke, with EoMT and a DINOv2-pretrained ViT-B, it reports 32.1 mAP versus ALGM’s 30.1, 153 versus 229 GFLOPs, and 142.8 versus 121.5 images per second. On Cityscapes, mAP is unchanged and throughput is similar.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.