Instance Segmentation on iSAID
Average Precision
- 2017-03-20 — Mask-RCNN+: Average Precision 37.18
- 2019-05-30 — PANet++: Average Precision 40.0
| Rank | Model | Average Precision | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | PANet++ | 40.00 | iSAID: A Large-scale Dataset for Instance Segmentation in Aerial Images | CAPTAIN-WHU/iSAID_Devkit · yeliudev/catnet · Anirudh0707/Roads-and-Building-Segmentation | 2019 |
| 2 | PANet+ | 39.54 | iSAID: A Large-scale Dataset for Instance Segmentation in Aerial Images | CAPTAIN-WHU/iSAID_Devkit · yeliudev/catnet · Anirudh0707/Roads-and-Building-Segmentation | 2019 |
| 3 | Mask-RCNN+ | 37.18 | Mask R-CNN | tensorflow/models · facebookresearch/detectron2 · facebookresearch/detectron · +176 | 2017 |
| 4 | Mask-RCNN | 36.50 | Mask R-CNN | tensorflow/models · facebookresearch/detectron2 · facebookresearch/detectron · +176 | 2017 |
| 5 | PANet | 34.17 | Path Aggregation Network for Instance Segmentation | ultralytics/yolov5 · open-mmlab/mmdetection · PaddlePaddle/PaddleDetection · +7 | 2018 |