martin-liao
commented
6 months ago Hello! Thanks for your interest in our work.
- As discussed in previous approaches ([7], [11], [12] in the paper), it is natural for humans to distinguish objects by perceiving both object body and shape (boundary) information. For example, it could be really hard to distinguish a bicycle/motorcycle and a bus/truck if the boundary information is unavailable. Therefore, we propose to fuse body and boundary information for the semantic segmentation task.
- Most canonical edge detection works like BDCN and RCF have pointed out that multi-scale information is crucial to handle different scales and levels edge. So, we extract multi-scale features to extract both low-level and high-level boundary information.
- We have been motivated by recent lightweight semantic segmentation and dual-task learning approaches. Furthermore, considering the boundary features could benefit the visual odometry and the semantic features could benefit the semantic mapping, we propose the Mobile-Seed for online SLAM tasks.
Billy-ZTB
Hello! I have had the opportunity to read your paper, in which you wrote, "... Therefore, the fusion weights should be conditioned on the input. There are dynamic fusion methods [17], [19] that can adapt the weights for semantic edge detection and semantic segmentation tasks.However, calculating fusion weights in both spatial and channel dimensions is still too cumbersome for the lightweight framework." While [17] represents the paper "Dynamic Feature Fusion for Semantic Edge Detection" , I am wondering how this "adaptive weight learner" proposed in the DFF model could fuse semantic edge detection and semantic segmentation results. I am guessing that maybe by inserting edge detection layers into semantic segmentation layers and multiply this tensor by the weight came from the "adaptive weight learner" could it help to fuse these two tasks' outputs. But fusing different side outputs of edge detection is so different from fusing the results of edge and segmentation, could it still be helpful? I would greatly appreciate your insights on this matter. By the way, the method proposed from your work is impressive, I am also curious about how this idea came up into your mind😊.