Reproduce “Generating Enhanced Negatives for Training Language-Based Object Detectors” 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
Shiyu Zhao, Long Zhao, Vijay Kumar B.G, Yumin Suh, Dimitris N. Metaxas, Manmohan Chandraker, Samuel Schulter
The authors improve language-based object detection by adding generated negative descriptions and images to detector training. They use language models to create related but contradictory descriptions and an image generator to edit corresponding regions, with filtering for noisy images. On D³, FIBER-B full-description AP rises from 22.7 to 26.0; image-negative benefits vary by benchmark and training setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.