use_gpu: true
log_iter: 5
save_dir: output
snapshot_epoch: 10
weights: output/tinypose_128x96/model_final
epoch: 420
num_joints: &num_joints 17
pixel_std: &pixel_std 200
metric: KeyPointTopDownCOCOEval
num_classes: 1
train_height: &train_height 128
train_width: &train_width 96
trainsize: &trainsize [*train_width, *train_height]
hmsize: &hmsize [24, 32]
flip_perm: &flip_perm [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12], [13, 14], [15, 16]]

# AMP training
init_loss_scaling: 32752
master_grad: true

#####model
architecture: TopDownHRNet

TopDownHRNet:
  backbone: LiteHRNet
  post_process: HRNetPostProcess
  flip_perm: *flip_perm
  num_joints: *num_joints
  width: &width 40
  loss: KeyPointMSELoss
  use_dark: true

LiteHRNet:
  network_type: wider_naive
  freeze_at: -1
  freeze_norm: false
  return_idx: [0]

KeyPointMSELoss:
  use_target_weight: true
  loss_scale: 1.0

#####optimizer
LearningRate:
  base_lr: 0.008
  schedulers:
  - !PiecewiseDecay
    milestones: [380, 410]
    gamma: 0.1
  - !LinearWarmup
    start_factor: 0.001
    steps: 500

OptimizerBuilder:
  optimizer:
    type: Adam
  regularizer:
    factor: 0.0
    type: L2


#####data
TrainDataset:
  !KeypointTopDownCocoDataset
    image_dir: ""
    anno_path: aic_coco_train_cocoformat.json
    dataset_dir: dataset
    num_joints: *num_joints
    trainsize: *trainsize
    pixel_std: *pixel_std
    use_gt_bbox: True


EvalDataset:
  !KeypointTopDownCocoDataset
    image_dir: val2017
    anno_path: annotations/person_keypoints_val2017.json
    dataset_dir: dataset/coco
    num_joints: *num_joints
    trainsize: *trainsize
    pixel_std: *pixel_std
    use_gt_bbox: True
    image_thre: 0.5

TestDataset:
  !ImageFolder
    anno_path: dataset/coco/keypoint_imagelist.txt

worker_num: 2
global_mean: &global_mean [0.485, 0.456, 0.406]
global_std: &global_std [0.229, 0.224, 0.225]
TrainReader:
  sample_transforms:
    - RandomFlipHalfBodyTransform:
        scale: 0.25
        rot: 30
        num_joints_half_body: 8
        prob_half_body: 0.3
        pixel_std: *pixel_std
        trainsize: *trainsize
        upper_body_ids: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
        flip_pairs: *flip_perm
    - AugmentationbyInformantionDropping:
        prob_cutout: 0.5
        offset_factor: 0.05
        num_patch: 1
        trainsize: *trainsize
    - TopDownAffine:
        trainsize: *trainsize
        use_udp: true
    - ToHeatmapsTopDown_DARK:
        hmsize: *hmsize
        sigma: 1
  batch_transforms:
    - NormalizeImage:
        mean: *global_mean
        std: *global_std
        is_scale: true
    - Permute: {}
  batch_size: 512
  shuffle: true
  drop_last: false

EvalReader:
  sample_transforms:
    - TopDownAffine:
        trainsize: *trainsize
        use_udp: true
  batch_transforms:
    - NormalizeImage:
        mean: *global_mean
        std: *global_std
        is_scale: true
    - Permute: {}
  batch_size: 16

TestReader:
  inputs_def:
    image_shape: [3, *train_height, *train_width]
  sample_transforms:
    - Decode: {}
    - TopDownEvalAffine:
        trainsize: *trainsize
    - NormalizeImage:
        mean: *global_mean
        std: *global_std
        is_scale: true
    - Permute: {}
  batch_size: 1
  fuse_normalize: false