dla.md

DLA

• Paper：Deep Layer Aggregation

• Origin Repo：ucbdrive/dla

• Code：dla.py

• Evaluate Transforms：

  # backend: pil
  # input_size: 224x224
  transforms = T.Compose([
      T.Resize(256, interpolation='bilinear'),
      T.CenterCrop(224),
      T.ToTensor(),
      T.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
  ])

• Model Details：

  Model	Model Name	Params (M)	FLOPs (G)	Top-1 (%)	Top-5 (%)	Pretrained Model
  DLA-34	dla_34	15.8	3.1	76.39	93.15	Download
  DLA-46-c	dla_46_c	1.3	0.5	64.88	86.29	Download
  DLA-46x-c	dla_46x_c	1.1	0.5	65.98	86.98	Download
  DLA-60	dla_60	22.0	4.2	77.02	93.31	Download
  DLA-60x	dla_60x	17.4	3.5	78.24	94.02	Download
  DLA-60x-c	dla_60x_c	1.3	0.6	67.91	88.43	Download
  DLA-102	dla_102	33.3	7.2	79.44	94.76	Download
  DLA-102x	dla_102x	26.4	5.9	78.51	94.23	Download
  DLA-102x2	dla_102x2	41.4	9.3	79.45	94.64	Download
  DLA-169	dla_169	53.5	11.6	78.71	94.34	Download

• Citation：

  @misc{yu2019deep,
      title={Deep Layer Aggregation}, 
      author={Fisher Yu and Dequan Wang and Evan Shelhamer and Trevor Darrell},
      year={2019},
      eprint={1707.06484},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
  }