YOLACT++

Review of paper named YOLACT++: Better Real-time Instance Segmentation

Above figure is the main reason of YOLACT++ is accepted in ICCV 2019. Most of the results in computer vision were focused on highering mAP score. However this was focused on real time instance segmentation by using only one gpu. Which I think that it is valuable to review and use this model into real industry. This research used single Titan Xp to make 34.1mAP on MS COCO at 33.5 fps.

Reference

[1] https://arxiv.org/abs/1912.06218

Github

[1] https://github.com/dbolya/yolact