config.yml

dataset: 
  # _name: pascal_voc_classification
  split: trainval
  data_dir: ./datasets/VOCdevkit
  year: 2012

train_transform:
  # _name: image_transform
  augmentation:
    horizontal_flip: 0.5
  image_size: [448, 448]
  mean: [0.485, 0.456, 0.406]
  std: [0.229, 0.224, 0.225]

test_transform:
  # _name: image_transform
  image_size: [448, 448]
  mean: [0.485, 0.456, 0.406]
  std: [0.229, 0.224, 0.225]

data_loaders:
  # _name: fetch_data
  batch_size: 16
  num_workers: 4
  train_augmentation:
    horizontal_flip: 0.5
  train_splits:
    - trainval

model:
  # _name: peak_response_mapping
  # backbone:
  #   _name: fc_resnet50
  win_size: 3
  sub_pixel_locating_factor: 8
  enable_peak_stimulation: true

optimizer:
  # _name: sgd_optimizer
  lr: 0.01
  momentum: 0.9
  weight_decay: 1.0e-4

finetune:
  # _name: finetune
  base_lr: 0.01
  groups:
    'features': 0.01

meters:
  loss:
    _name: loss_meter

max_epoch: 20

device: cuda
log: ./

snapshot: ./snapshots