Reproduce “Two-Stream Multi-Task Network for Fashion Recognition” 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
Peizhao Li, Yanjing Li, Xiaolong Jiang, Xiantong Zhen
The authors present a fashion-image system that predicts garment category, attributes, and landmarks together. An hourglass-based landmark stream guides a ResNet50 classification stream through a pixel-level boundary attention map and shared structural features. On DeepFashion-C, its top-5 attribute recall is 77.91, versus 60.95 for Wang et al.’s method; this is a benchmark result, not a runtime or deployment measure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.