Semantic Segmentation on COCO-Stuff 10K (mIoU SS)
53.2mIoU (SS)ViT-Adapter
Evaluation results
| Method | Links | |
|---|---|---|
| ViT-Adapterframework=mask cls, backbone=BEIT-L, PRM=568M, R/P=87%, test time=342ms, batch size=1, device=RTX 30902023.10 | 53.2 | |
| PlainSegframework=mask cls, backbone=BEIT-L, PRM=332M, R/P=10%, test time=195ms(-43%), batch size=1, device=RTX 30902023.10 | 53.02 | |
| PlainSeg-Hierframework=mask cls, backbone=BEIT-L, PRM=326M, R/P=8%, test time=188ms(-45%), batch size=1, device=RTX 30902023.10 | 52.99 | |
| PlainSegframework=mask cls, backbone=BEIT-B, PRM=105M, R/P=22%, test time=107ms(-28%), batch size=1, device=RTX 30902023.10 | 51.09 | |
| PlainSeg-Hierframework=mask cls, backbone=BEIT-B, PRM=105M, R/P=22%, test time=107ms(-28%), batch size=1, device=RTX 30902023.10 | 51.01 | |
| ViT-Adapterframework=mask cls, backbone=BEIT-B, PRM=120M, R/P=40%, test time=149ms, batch size=1, device=RTX 30902023.10 | 50 |