Reproduce “Rethinking Infrastructure Inspection as Image Difference Classification: A Traffic Sign Case Study” 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
Ching Yau Fergus Mok, Lavindra de Silva, Varun Kumar Reja, Ioannis Brilakis
The authors test whether an earlier, undamaged image of the same road sign helps classify its current condition when annotated data are scarce. They compare a fine-tuned instruction-based Qwen3-VL-8B classifier using the reference with a matched single-image pipeline. Macro-F1 rises from 0.511 to 0.544 at 1-shot and 0.563 to 0.601 at 8-shot; gains were significant only at those settings, not across models generally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.