Collection

To keep related model together, you can create Collections. The metadata format for collections is the same as for models and all member models inherit all the metadata and can override/add to it.

The fields of a collection are exactly the same as fields of a Model - see the description there.

A full example

Collections:
  - Name: Mask R-CNN
    Metadata:
      Training Data: COCO
      Training Techniques: 
        - SGD with Momentum
      Training Resources: 8x V100 GPUs
      Architecture:
        - RoI Align
        - RPN
    Paper: https://arxiv.org/abs/1703.06870v3
    README: docs/maskrcnn.md

Models:
  - Name: Mask R-CNN (R101-C4, 3x)
    In Collection: Mask R-CNN 
    Metadata:
      inference time (s/im): 0.145
      train time (s/iter): 0.652
      Training Memory (GB): 6.3
    Results:
      - Task: Object Detection
        Dataset: COCO minival
        Metrics:
          box AP: 42.6
      - Task: Instance Segmentation
        Dataset: COCO minival
        Metrics:
          mask AP: 36.7
    Weights: https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_101_C4_3x/138363239/model_final_a2914c.pkl
  
  - Name: Mask R-CNN (R50-C4, 3x)
    In Collection: Mask R-CNN 
    Metadata:
      inference time (s/im): 0.111
      train time (s/iter): 0.575
      Training Memory (GB): 5.2
    Results:
      - Task: Object Detection
        Dataset: COCO minival
        Metrics:
          box AP: 39.8
      - Task: Instance Segmentation
        Dataset: COCO minival
        Metrics:
          mask AP: 34.4
    Weights: https://dl.fbaipublicfiles.com/detectron2/COCO-InstanceSegmentation/mask_rcnn_R_50_C4_3x/137849525/model_final_4ce675.pkl

In this example we have two variants of the Mask R-CNN model, one with a ResNet-50 backbone and one with a ResNet-101 backbone. These models belong together as variants of the Mask R-CNN, so we link them via a Mask R-CNN model collection.