hrnet_w32_384x288.yml

use_gpu: true
log_iter: 5
save_dir: output
snapshot_epoch: 10
weights: output/hrnet_w32_384x288/model_final
epoch: 210
num_joints: &num_joints 17
pixel_std: &pixel_std 200
metric: KeyPointTopDownCOCOEval
num_classes: 1
train_height: &train_height 384
train_width: &train_width 288
trainsize: &trainsize [*train_width, *train_height]
hmsize: &hmsize [72, 96]
flip_perm: &flip_perm [[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12], [13, 14], [15, 16]]


#####model
architecture: TopDownHRNet
pretrain_weights: https://paddledet.bj.bcebos.com/models/pretrained/Trunc_HRNet_W32_C_pretrained.pdparams

TopDownHRNet:
  backbone: HRNet
  post_process: HRNetPostProcess
  flip_perm: *flip_perm
  num_joints: *num_joints
  width: &width 32
  loss: KeyPointMSELoss
  flip: true

HRNet:
  width: *width
  freeze_at: -1
  freeze_norm: false
  return_idx: [0]

KeyPointMSELoss:
  use_target_weight: true


#####optimizer
LearningRate:
  base_lr: 0.0005
  schedulers:
  - !PiecewiseDecay
    milestones: [170, 200]
    gamma: 0.1
  - !LinearWarmup
    start_factor: 0.001
    steps: 1000

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


#####data
TrainDataset:
  !KeypointTopDownCocoDataset
    image_dir: train2017
    anno_path: annotations/person_keypoints_train2017.json
    dataset_dir: dataset/coco
    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
    bbox_file: bbox.json
    num_joints: *num_joints
    trainsize: *trainsize
    pixel_std: *pixel_std
    use_gt_bbox: True
    image_thre: 0.0


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: 0.5
        rot: 40
        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
    - TopDownAffine:
        trainsize: *trainsize
    - ToHeatmapsTopDown:
        hmsize: *hmsize
        sigma: 2
  batch_transforms:
    - NormalizeImage:
        mean: *global_mean
        std: *global_std
        is_scale: true
    - Permute: {}
  batch_size: 64
  shuffle: true
  drop_last: false

EvalReader:
  sample_transforms:
    - TopDownAffine:
        trainsize: *trainsize
  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