Search papers, labs, and topics across Lattice.
This paper introduces a ResNet-50 based pipeline for multi-label gastrointestinal video analysis, predicting 17 anatomical and pathological labels. To combat severe class imbalance, they employed clipped class-wise positive weighting in the loss function, enhancing rare-class learning. They further refined temporal event decoding by integrating ground truth-style framewise event composition, anatomy vote smoothing, and anatomy-based pathology gating, achieving a significant improvement in temporal mAP.
Class reweighting and anatomy-guided decoding can substantially improve the performance of video analysis pipelines for rare events in imbalanced gastrointestinal datasets.
We developed a multi-label gastrointestinal video analysis pipeline based on a ResNet-50 frame classifier followed by anatomy-guided temporal event decoding. The system predicts 17 labels, including 5 anatomy classes and 12 pathology classes, from frames resized to 336x336. A major challenge was severe class imbalance, particularly for rare pathology labels. To address this, we used clipped class-wise positive weighting in the training loss, which improved rare-class learning while maintaining stable optimization. At the temporal stage, we found that direct frame-to-event conversion produced fragmented mismatches with the official ground truth. The final submission therefore combined GT-style framewise event composition, anatomy vote smoothing, and anatomy-based pathology gating with a conservative hysteresis decoder. This design improved the final temporal mAP from 0.3801 to 0.4303 on the challenge test set.
