Instance Segmentation on NYU v2 (test)
62.64PQEMSANet
Evaluation results
| Method | Links | |||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EMSANetBackbone=2x ResNet101, Task(s)=Sem + Sce + Ins + Or2023.06 | 62.64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMSANetBackbone=2x ResNet34-NBt1D, Task(s)=Sem + Sce + Ins + Or2023.06 | 61.33 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMSAFormerBackbone=SwinV2-T-128-Multi-Aug, Task(s)=Sem + Sce + Ins + Or, Semantic Decoder=EMSANet2023.06 | 59.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMSAFormerBackbone=SwinV2-T-128-Multi-Aug, Task(s)=Sem(SegFormer) + Sce + Ins + Or, Semantic Decoder=SegFormer2023.06 | 58.75 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| EMSAFormer (SwinV2-T-128-Multi-Aug single-task)Backbone=SwinV2-T-128-Multi-Aug, Task(s)=Instance Segmentation (Ins), Throughput (50W)=48.4 FPS, Throughput (30W)=33.5 FPS2023.06 | 56.44 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seg-Cluster2017.11 | — | 23.2 | 31 | 70.1 | 27.1 | 1.3 | 25.8 | 20.3 | 13.9 | 11.1 | 24.3 | 4.4 | 16.3 | 3.6 | 32 | 25.5 | 36.3 | 50.9 | 12.9 | 2.2 | 23.5 | |
| SGPN2017.11 | — | 26.5 | 55.9 | 53.3 | 27.8 | 0 | 27.4 | 59.6 | 28.9 | 6.1 | 33.9 | 2 | 19.7 | 2 | 29.4 | 30.7 | 39.1 | 43.6 | 17.6 | 1.2 | 25.9 | |
| SGPN-CNNCNN features=Incorporated 2D features2017.11 | — | 30.5 | 56.4 | 55.4 | 35.2 | 0 | 42.6 | 50.6 | 23.1 | 21.1 | 31.8 | 7.5 | 22.7 | 6.4 | 39.9 | 33.5 | 42.4 | 54.8 | 21.3 | 3.8 | 32.1 |