ReCal-Net: Semantic Segmentation in Cataract Surgery Videos
Our paper has been accepted for publication at the ICONIP 2021 conference.
Title: ReCal-Net: Joint Region-Channel-Wise Calibrated Network for Semantic Segmentation in Cataract Surgery Videos
Authors: Negin Ghamsarian, Mario Taschwer, Doris Putzgruber-Adamitsch, Stephanie Sarny, Yosuf El-Shabrawi, Klaus Schöffmann
Abstract: Semantic segmentation in surgical videos is a prerequisite for a broad range of applications towards improving surgical outcomes and surgical video analysis. However, semantic segmentation in surgical videos involves many challenges. In particular, in cataract surgery, various features of the relevant objects such as blunt edges, color and context variation, reflection, transparency, and motion blur pose a challenge for semantic segmentation. In this paper, we propose a novel convolutional module termed as ReCal module, which can calibrate the feature maps by employing region and channel inter-dependencies and cross-dependencies. This calibration strategy can effectively enhance semantic representation by correlating different representations of the same semantic label, considering a multi-angle local view centering around each pixel. Thus the proposed module can deal with distant visual characteristics of unique objects as well as cross-similarities in the visual characteristics of different objects. Moreover, we propose a novel network architecture based on the proposed module termed as ReCal-Net. Experimental results confirm the superiority of ReCal-Net compared to rival state-of-the-art approaches for all relevant objects in cataract surgery. Moreover, ablation studies reveal the effectiveness of the ReCal module in boosting semantic segmentation accuracy.