youngwanLEE/centermask2
CenterMask2 is a real-time anchor-free instance segmentation model based on detectron2/PyTorch.

Velocity · 7d
+0.3
★ / day
Trend
→steady
star history
This repository implements CenterMask2, an upgraded instance segmentation model published at CVPR 2020. It performs real-time anchor-free instance segmentation on top of detectron2 and PyTorch, offering both full-scale and lightweight CenterMask-Lite variants for different speed/accuracy tradeoffs. The model uses VoVNetV2 as its backbone network and achieves state-of-the-art results on COCO benchmark outperforming methods like Mask R-CNN and YOLACT++.