mask_rcnn_r50_vd_fpn_1x_coco.yml

_BASE_: [
  'mask_rcnn_r50_fpn_1x_coco.yml',
]

pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/ResNet50_vd_pretrained.pdparams
weights: output/mask_rcnn_r50_vd_fpn_1x_coco/model_final

ResNet:
  # index 0 stands for res2
  depth: 50
  variant: d
  norm_type: bn
  freeze_at: 0
  return_idx: [0,1,2,3]
  num_stages: 4