CamoNAS Boosts Camouflaged Object Detection with NAS

Camouflaged Object Detection (COD) is a challenging computer vision task focused on identifying and segmenting objects that blend seamlessly into their surroundings, often characterized by weak edge cues and indistinct boundaries. Traditional COD models typically rely on manually designed architectures and multi-scale feature fusion, which are often based on heuristics rather than systematic optimization. Researchers introduce CamoNAS, a novel framework that employs Neural Architecture Search (NAS) to automatically discover highly effective architectures for COD. CamoNAS explores a hierarchical search space, optimizing both cell-level operations and network-level downsampling paths, specifically tailored for detecting camouflaged objects. A key innovation is its RGB frequency dual-stream architecture, which integrates a learnable wavelet transform to complement the standard RGB spatial stream, enhancing the model's ability to discern subtle patterns. CamoNAS has achieved state-of-the-art performance across four major COD benchmarks, demonstrating the significant advantages of applying NAS to this complex detection problem. The code is open-sourced.