FishNet: One Stage to Detect, Segmentation and Pose Estimation

In this project, we combine target detection, instance segmentation, and keypoint detection into the one network structure and use a YOLO-like model to give predicted values using one phase. We use the idea of multi-task network to merge the three tasks, and use the labeled fish images as the experimental objects, and finally achieve good results.

Update Later...

We have now succeeded in converging the network and obtained good results, more optimized hyperparameters are being debugged, the specific results will be released after the paper is accepted The prediction diagram of the experimental procedure in ". /experiments_image/"