blazeface.yml 2.7 KB
Newer Older
1 2
architecture: BlazeFace
max_iters: 320000
3
pretrain_weights:
4 5 6 7 8
use_gpu: true
snapshot_iter: 10000
log_iter: 20
metric: WIDERFACE
save_dir: output
W
wangguanzhong 已提交
9
weights: output/blazeface/model_final
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
# 1(label_class) + 1(background)
num_classes: 2

BlazeFace:
  backbone: BlazeNet
  output_decoder:
    keep_top_k: 750
    nms_threshold: 0.3
    nms_top_k: 5000
    score_threshold: 0.01
  min_sizes: [[16.,24.], [32., 48., 64., 80., 96., 128.]]
  use_density_prior_box: false

BlazeNet:
  with_extra_blocks: true
  lite_edition: false

LearningRate:
  base_lr: 0.001
  schedulers:
  - !PiecewiseDecay
    gamma: 0.1
    milestones: [240000, 300000]

OptimizerBuilder:
  optimizer:
    momentum: 0.0
    type: RMSPropOptimizer
  regularizer:
    factor: 0.0005
    type: L2

42 43 44 45
TrainReader:
  inputs_def:
    image_shape: [3, 640, 640]
    fields: ['image', 'gt_bbox', 'gt_class']
46
  dataset:
47
    !WIDERFaceDataSet
48
    dataset_dir: dataset/wider_face
49
    anno_path: wider_face_split/wider_face_train_bbx_gt.txt
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
    image_dir: WIDER_train/images
  sample_transforms:
  - !DecodeImage
    to_rgb: true
  - !NormalizeBox {}
  - !RandomDistort
    brightness_lower: 0.875
    brightness_upper: 1.125
    is_order: true
  - !ExpandImage
    max_ratio: 4
    prob: 0.5
  - !CropImageWithDataAchorSampling
    anchor_sampler:
    - [1, 10, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.2, 0.0]
    batch_sampler:
    - [1, 50, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0]
    - [1, 50, 0.3, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0]
    - [1, 50, 0.3, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0]
    - [1, 50, 0.3, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0]
    - [1, 50, 0.3, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0]
    target_size: 640
  - !RandomInterpImage
    target_size: 640
  - !RandomFlipImage
    is_normalized: true
  - !Permute {}
  - !NormalizeImage
    is_scale: false
    mean: [104, 117, 123]
    std: [127.502231, 127.502231, 127.502231]
81 82
  batch_size: 8
  use_process: true
83
  worker_num: 8
84
  shuffle: true
85

86 87 88
EvalReader:
  inputs_def:
    fields: ['image', 'im_id']
89
  dataset:
90
    !WIDERFaceDataSet
91
    dataset_dir: dataset/wider_face
92
    anno_path: wider_face_split/wider_face_val_bbx_gt.txt
93 94 95 96 97 98
    image_dir: WIDER_val/images
  sample_transforms:
  - !DecodeImage
    to_rgb: true
  - !NormalizeBox {}
  - !NormalizeImage
99
    is_channel_first: false
100
    is_scale: false
101
    mean: [123, 117, 104]
102
    std: [127.502231, 127.502231, 127.502231]
103
  - !Permute {}
104
  batch_size: 1
105 106 107 108

TestReader:
  inputs_def:
    fields: ['image', 'im_id', 'im_shape']
109
  dataset:
110
    !ImageFolder
111 112 113 114 115
    use_default_label: true
  sample_transforms:
  - !DecodeImage
    to_rgb: true
  - !NormalizeImage
116
    is_channel_first: false
117
    is_scale: false
118
    mean: [123, 117, 104]
119
    std: [127.502231, 127.502231, 127.502231]
120
  - !Permute {}
121
  batch_size: 1