Reproduce “CrackSeg9k: A Collection and Benchmark for Crack Segmentation Datasets and Frameworks” 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
Shreyas Kulkarni, Shreyas Singh, Dhananjay Balakrishnan, Siddharth Sharma, Saipraneeth Devunuri, Sai Chowdeswara Rao Korlapati
CrackSeg9k combines existing crack-image datasets and refines selected masks to provide a shared benchmark for pixel-level crack segmentation. The authors evaluate segmentation models on its reported test split and use DINO attention features as an additional input. Their DeepLab V3+ model reports F1 0.7238 and mIoU 0.7712, compared with DeepCrack’s 0.3404 and 0.5813; glass-like backgrounds remain difficult.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.