Fig. 3
From: ISSEC: inferring contacts among protein secondary structure elements using deep object detection
Deep object detection model used by ISSEC. (i) Feature extractor aims to extract high-level features from the predicted probabilities of inter-residue contacts. (ii) Region proposing network aims to propose candidate rectangle regions that might contain contacting SSEs. (iii) Mask-RCNN aims to judge whether a candidate rectangle region contains characteristic patterns or not