deep learning for bleeding detection in endoscopic capsule images

this paper discusses an algorithm for detecting bleeding in images taken from an endoscopic capsule. this algorithm consists of two parts. first, with deep learning, they instruct a deep network to distinguish between blood images and normal images. then the images in the blood class are transmitted to the second part of the algorithm. in the second part, the images are converted to hsv and by comparing each pixel with the threshold of blood, the location of the bleeding is marked and indicated by a green rectangle. simulation of this algorithm is implemented using the python language and tensorflow. the results indicate that the deep network has been able to categorize well between blood images and normal images, and the location of bleeding is also prominently indicated.