Camouflage Object Detection System: A Review

Even for the unaided eye, it is difficult to discern camouflaged items in their natural context. CamoNet, an unique bio-inspired network that utilizes both instance segmentation and a bio-inspired attack stream, is proposed for camouflaged object segmentation. Our proposed network, unlike existing segmentation networks, contains two segmentation streams: the main stream and the bio-inspired attack stream, which correlate to the original and flipped pictures, respectively. To increase segmentation accuracy, the output of the bio-inspired assault stream is combined with the result of the mainstream for the final camouflage map. Furthermore, before using an edge detection approach, our method applies an image enhancement technique to improve the underlit and foggy input image. We’re also looking for more precise characteristics than simple texture-based features to increase the accuracy of spotting disguised objects.