metafile.yml

Collections:
  - Name: Conformer
    Metadata:
      Training Data: ImageNet-1k
      Architecture:
        - Layer Normalization
        - Scaled Dot-Product Attention
        - Dropout
    Paper:
      URL: https://arxiv.org/abs/2105.03889
      Title: "Conformer: Local Features Coupling Global Representations for Visual Recognition"
    README: configs/conformer/README.md
    Code:
      URL: https://github.com/open-mmlab/mmpretrain/blob/v0.19.0/mmcls/models/backbones/conformer.py
      Version: v0.19.0

Models:
  - Name: conformer-tiny-p16_3rdparty_in1k
    In Collection: Conformer
    Config: configs/conformer/conformer-tiny-p16_8xb128_in1k.py
    Metadata:
      FLOPs: 4899611328
      Parameters: 23524704
    Results:
      - Dataset: ImageNet-1k
        Metrics:
          Top 1 Accuracy: 81.31
          Top 5 Accuracy: 95.60
        Task: Image Classification
    Weights: https://download.openmmlab.com/mmclassification/v0/conformer/conformer-tiny-p16_3rdparty_8xb128_in1k_20211206-f6860372.pth
    Converted From:
      Weights: https://drive.google.com/file/d/19SxGhKcWOR5oQSxNUWUM2MGYiaWMrF1z/view?usp=sharing
      Code: https://github.com/pengzhiliang/Conformer/blob/main/models.py#L65
  - Name: conformer-small-p16_3rdparty_in1k
    In Collection: Conformer
    Config: configs/conformer/conformer-small-p16_8xb128_in1k.py
    Metadata:
      FLOPs: 10311309312
      Parameters: 37673424
    Results:
      - Dataset: ImageNet-1k
        Metrics:
          Top 1 Accuracy: 83.32
          Top 5 Accuracy: 96.46
        Task: Image Classification
    Weights: https://download.openmmlab.com/mmclassification/v0/conformer/conformer-small-p16_3rdparty_8xb128_in1k_20211206-3065dcf5.pth
    Converted From:
      Weights: https://drive.google.com/file/d/1mpOlbLaVxOfEwV4-ha78j_1Ebqzj2B83/view?usp=sharing
      Code: https://github.com/pengzhiliang/Conformer/blob/main/models.py#L73
  - Name: conformer-small-p32_8xb128_in1k
    In Collection: Conformer
    Config: configs/conformer/conformer-small-p32_8xb128_in1k.py
    Metadata:
      FLOPs: 7087281792
      Parameters: 38853072
    Results:
      - Dataset: ImageNet-1k
        Metrics:
          Top 1 Accuracy: 81.96
          Top 5 Accuracy: 96.02
        Task: Image Classification
    Weights: https://download.openmmlab.com/mmclassification/v0/conformer/conformer-small-p32_8xb128_in1k_20211206-947a0816.pth
  - Name: conformer-base-p16_3rdparty_in1k
    In Collection: Conformer
    Config: configs/conformer/conformer-base-p16_8xb128_in1k.py
    Metadata:
      FLOPs: 22892078080
      Parameters: 83289136
    Results:
      - Dataset: ImageNet-1k
        Metrics:
          Top 1 Accuracy: 83.82
          Top 5 Accuracy: 96.59
        Task: Image Classification
    Weights: https://download.openmmlab.com/mmclassification/v0/conformer/conformer-base-p16_3rdparty_8xb128_in1k_20211206-bfdf8637.pth
    Converted From:
      Weights: https://drive.google.com/file/d/1oeQ9LSOGKEUaYGu7WTlUGl3KDsQIi0MA/view?usp=sharing
      Code: https://github.com/pengzhiliang/Conformer/blob/main/models.py#L89