Segmentation Model Zoo
February 17, 2023 ยท View on GitHub
FCN
Pretrained on Pascal VOC 2012 + Aug.
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(V100) (ms/img) | mIoU | Download |
|---|---|---|---|---|---|---|
| fcn_r50_d8 | fcn_r50-d8_512x512_8xb4_60e_voc12aug | 23M/49M | 19.8 | 166ms | 69.01 | model - log |
UperNet
Pretrained on Pascal VOC 2012 + Aug.
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(V100) (ms/img) | mIoU | Download |
|---|---|---|---|---|---|---|
| upernet_r50 | upernet_r50_512x512_8xb4_60e_voc12aug | 23M/66M | 5.5 | 282.9ms | 76.59 | model - log |
STDC
trained on Cityscapes.
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(V100) (ms/img) | mIoU | Download |
|---|---|---|---|---|---|---|
| STDC1 | stdc1_cityscape_8xb6_e1290 | 7.7M/8.5M | 4.5 | 11.9ms | 75.4 | model - log |
| STDC2 | stdc2_cityscape_8xb6_e1290 | 11.6M/12.6M | 5.6 | 15.4ms | 76.6 | model - log |
Mask2former
Instance Segmentation on COCO
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(A100) (ms/img) | box MAP | Mask mAP | Download |
|---|---|---|---|---|---|---|---|
| mask2former_r50 | mask2former_r50_8xb2_e50_instance | 23.5M/44M | 18.8 | 214ms | 46.09 | 43.26 | model - log |
Panoptic Segmentation on COCO
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(A100) (ms/img) | PQ | box MAP | Mask mAP | Download |
|---|---|---|---|---|---|---|---|---|
| mask2former_r50 | mask2former_r50_8xb2_e50_panopatic | 23.5M/44M | 18.8 | 241ms | 51.64 | 44.81 | 41.88 | model - log |
Semantic Segmentation on ADE20K
| Algorithm | Config | Params (backbone/total) | Train memory (GB) | inference time(A100) (ms/img) | mIOU | Download |
|---|---|---|---|---|---|---|
| mask2former_r50 | mask2former_r50_8xb2_e127_semantic | 23.5M/44M | 5.6 | 504ms | 47.03 | model - log |
SegFormer
Semantic segmentation models trained on CoCo_stuff164k.
| Algorithm | Config | Params (backbone/total) | inference time(V100) (ms/img) | mIoU | Download |
|---|---|---|---|---|---|
| SegFormer_B0 | segformer_b0_coco.py | 3.3M/3.8M | 47.2ms | 35.91 | model - log |
| SegFormer_B1 | segformer_b1_coco.py | 13.2M/13.7M | 46.8ms | 40.53 | model - log |
| SegFormer_B2 | segformer_b2_coco.py | 24.2M/27.5M | 49.1ms | 44.53 | model - log |
| SegFormer_B3 | segformer_b3_coco.py | 44.1M/47.4M | 52.3ms | 45.49 | model - log |
| SegFormer_B4 | segformer_b4_coco.py | 60.8M/64.1M | 58.5ms | 46.27 | model - log |
| SegFormer_B5 | segformer_b5_coco.py | 81.4M/85.7M | 99.2ms | 46.75 | model - log |