Reproduce “Scale Match for Tiny Person Detection” 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
Xuehui Yu, Yuqi Gong, Nan Jiang, Qixiang Ye, Zhenjun Han
The authors reshape pretraining image scales to match TinyPerson’s object-size distribution before training a detector for distant maritime person detection. Monotone Scale Match resizes whole source images while preserving object-size ordering. With Faster R-CNN-FPN, AP for 2–20-pixel objects at IoU 0.5 is 47.29%, versus 43.03% for mean-size-matched COCO100, an increase of 4.26 AP points in this experiment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.