Collections:

- Name: emanet

  Metadata:

    Training Data:

    - Cityscapes

  Paper:

    URL: https://arxiv.org/abs/1907.13426

    Title: Expectation-Maximization Attention Networks for Semantic Segmentation

  README: configs/emanet/README.md

  Code:

    URL: https://github.com/open-mmlab/mmsegmentation/blob/v0.17.0/mmseg/models/decode_heads/ema_head.py#L80

    Version: v0.17.0

  Converted From:

    Code: https://xialipku.github.io/EMANet

Models:

- Name: emanet_r50-d8_512x1024_80k_cityscapes

  In Collection: emanet

  Metadata:

    backbone: R-50-D8

    crop size: (512,1024)

    lr schd: 80000

    inference time (ms/im):

    - value: 218.34

      hardware: V100

      backend: PyTorch

      batch size: 1

      mode: FP32

      resolution: (512,1024)

    Training Memory (GB): 5.4

  Results:

  - Task: Semantic Segmentation

    Dataset: Cityscapes

    Metrics:

      mIoU: 77.59

      mIoU(ms+flip): 79.44

  Config: configs/emanet/emanet_r50-d8_512x1024_80k_cityscapes.py

  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/emanet/emanet_r50-d8_512x1024_80k_cityscapes/emanet_r50-d8_512x1024_80k_cityscapes_20200901_100301-c43fcef1.pth

- Name: emanet_r101-d8_512x1024_80k_cityscapes

  In Collection: emanet

  Metadata:

    backbone: R-101-D8

    crop size: (512,1024)

    lr schd: 80000

    inference time (ms/im):

    - value: 348.43

      hardware: V100

      backend: PyTorch

      batch size: 1

      mode: FP32

      resolution: (512,1024)

    Training Memory (GB): 6.2

  Results:

  - Task: Semantic Segmentation

    Dataset: Cityscapes

    Metrics:

      mIoU: 79.1

      mIoU(ms+flip): 81.21

  Config: configs/emanet/emanet_r101-d8_512x1024_80k_cityscapes.py

  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/emanet/emanet_r101-d8_512x1024_80k_cityscapes/emanet_r101-d8_512x1024_80k_cityscapes_20200901_100301-2d970745.pth

- Name: emanet_r50-d8_769x769_80k_cityscapes

  In Collection: emanet

  Metadata:

    backbone: R-50-D8

    crop size: (769,769)

    lr schd: 80000

    inference time (ms/im):

    - value: 507.61

      hardware: V100

      backend: PyTorch

      batch size: 1

      mode: FP32

      resolution: (769,769)

    Training Memory (GB): 8.9

  Results:

  - Task: Semantic Segmentation

    Dataset: Cityscapes

    Metrics:

      mIoU: 79.33

      mIoU(ms+flip): 80.49

  Config: configs/emanet/emanet_r50-d8_769x769_80k_cityscapes.py

  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/emanet/emanet_r50-d8_769x769_80k_cityscapes/emanet_r50-d8_769x769_80k_cityscapes_20200901_100301-16f8de52.pth

- Name: emanet_r101-d8_769x769_80k_cityscapes

  In Collection: emanet

  Metadata:

    backbone: R-101-D8

    crop size: (769,769)

    lr schd: 80000

    inference time (ms/im):

    - value: 819.67

      hardware: V100

      backend: PyTorch

      batch size: 1

      mode: FP32

      resolution: (769,769)

    Training Memory (GB): 10.1

  Results:

  - Task: Semantic Segmentation

    Dataset: Cityscapes

    Metrics:

      mIoU: 79.62

      mIoU(ms+flip): 81.0

  Config: configs/emanet/emanet_r101-d8_769x769_80k_cityscapes.py

  Weights: https://download.openmmlab.com/mmsegmentation/v0.5/emanet/emanet_r101-d8_769x769_80k_cityscapes/emanet_r101-d8_769x769_80k_cityscapes_20200901_100301-47a324ce.pth