Candidate Constrained CRFs for Loss-Aware Structured Prediction
This addresses the challenge of optimizing vision systems for specific metrics in tasks like image segmentation, though it is incremental as it builds on existing pipeline and CRF methods.
The paper tackled the problem of tuning computer vision systems to task-specific evaluation measures like Intersection-Over-Union, which top systems often ignore despite existing loss-aware prediction techniques. The result was a method that combined a highly tuned pipeline system with a CRF to enable loss-aware prediction, improving the pipeline's performance.
When evaluating computer vision systems, we are often concerned with performance on a task-specific evaluation measure such as the Intersection-Over-Union score used in the PASCAL VOC image segmentation challenge. Ideally, our systems would be tuned specifically to these evaluation measures. However, despite much work on loss-aware structured prediction, top performing systems do not use these techniques. In this work, we seek to address this problem, incorporating loss-aware prediction in a manner that is amenable to the approaches taken by top performing systems. Our main idea is to simultaneously leverage two systems: a highly tuned pipeline system as is found on top of leaderboards, and a traditional CRF. We show how to combine high quality candidate solutions from the pipeline with the probabilistic approach of the CRF that is amenable to loss-aware prediction. The result is that we can use loss-aware prediction methodology to improve performance of the highly tuned pipeline system.