HarDNet-MSEG is a new convolution neural network for polyp segmentation. It consists of a backbone and a decoder. The backbone is a low memory traffic CNN called HarDNet68, which has been successfully applied to various CV tasks including image classification, object detection, multi-object tracking, and semantic segmentation. The decoder part is inspired by the Cascaded Partial Decoder, which is known for fast and accurate salient object detection. The following image shows the result of Polyp Segmentation.
Figure 1. Polyp Segmentation Example