Camouflaged Object Detection
Overview
Camouflaged object detection seeks to locate and segment objects that resemble their backgrounds, where weak boundaries and misleading texture make ordinary visual cues unreliable. Methods use multi-scale features, edge and frequency cues, iterative refinement, and complementary inputs such as depth, thermal, polarization, video motion, or language; foundation models such as SAM are also adapted for supervised, weakly supervised, and training-free settings. Evaluation spans image and video benchmarks, with growing attention to open-vocabulary queries, corrupted images, missing modalities, and negative scenes where no target is present. A central challenge is improving boundary accuracy and generalization without increasing false positives, data dependence, or computational cost.
Latest papers
46 papers2 papers in the last four weeks, down 50% on the four weeks before. 0.0% of all new papers.