Camouflaged Object Detection in Images using a Search-Identification based framework

While identifying an object in an image is almost an instantaneous task for the human visual perception system, it takes more effort and time to process and identify a camouflaged object - an entity that flawlessly blends with the background in the image. This explains why it is much more challenging to enable a machine learning model to do the same, in comparison to generic object detection or salient object detection.

This project implements a framework called Search Identification Network, that simulates the search and identification pattern adopted by predators in hunting their prey and applies it to detect camouflaged objects. The efficiency of this framework in detecting polyps in medical image datasets is also measured.