CamoNAS Automates Architecture Search for Camouflaged Object Detection

CamoNAS uses a frequency-aware multi-resolution Neural Architecture Search to automatically optimize cell-level operations and downsampling paths for camouflaged object detection. The framework combines a learnable wavelet transform with an RGB spatial stream to achieve state-of-the-art results on four benchmarks.