DocumentAIKit/configs/smrt/ppyolo/ppyolov2_r50vd_dcn_365e_battery_1024.yml

architecture: YOLOv3
pretrain_weights: https://paddledet.bj.bcebos.com/models/ppyolov2_r50vd_dcn_365e_coco.pdparams
norm_type: sync_bn
use_ema: true
ema_decay: 0.9998
use_gpu: true
use_xpu: false
log_iter: 100
save_dir: output

metric: COCO
num_classes: 45

TrainDataset:
  !COCODataSet
    image_dir: images
    anno_path: annotations/train.json
    dataset_dir: dataset/battery_mini
    data_fields: ['image', 'gt_bbox', 'gt_class', 'is_crowd']

EvalDataset:
  !COCODataSet
    image_dir: images
    anno_path: annotations/test.json
    dataset_dir: dataset/battery_mini

TestDataset:
  !ImageFolder
    anno_path: annotations/test.json # also support txt (like VOC's label_list.txt)
    dataset_dir: dataset/battery_mini # if set, anno_path will be 'dataset_dir/anno_path'


epoch: 40
LearningRate:
  base_lr: 0.0001
  schedulers:
  - !PiecewiseDecay
    gamma: 0.1
    milestones:
    - 80
  - !LinearWarmup
    start_factor: 0.
    steps: 1000

snapshot_epoch: 5
worker_num: 2
TrainReader:
  inputs_def:
    num_max_boxes: 100
  sample_transforms:
    - Decode: {}
    - RandomDistort: {}
    - RandomFlip: {}
  batch_transforms:
    - BatchRandomResize: {target_size: [960, 992, 1024, 1056, 1088], random_size: True, random_interp: True, keep_ratio: False}
    - NormalizeBox: {}
    - PadBox: {num_max_boxes: 100}
    - BboxXYXY2XYWH: {}
    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
    - Permute: {}
    - Gt2YoloTarget: {anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]], anchors: [[10, 13], [16, 30], [33, 23], [30, 61], [62, 45], [59, 119], [116, 90], [156, 198], [373, 326]], downsample_ratios: [32, 16, 8]}
  batch_size: 4
  shuffle: true
  drop_last: true
  use_shared_memory: true

EvalReader:
  sample_transforms:
    - Decode: {}
    - Resize: {target_size: [1024, 1024], keep_ratio: False, interp: 2}
    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
    - Permute: {}
  batch_size: 8

TestReader:
  inputs_def:
    image_shape: [3, 1024, 1024]
  sample_transforms:
    - Decode: {}
    - Resize: {target_size: [1024, 1024], keep_ratio: False, interp: 2}
    - NormalizeImage: {mean: [0.485, 0.456, 0.406], std: [0.229, 0.224, 0.225], is_scale: True}
    - Permute: {}
  batch_size: 1


OptimizerBuilder:
  clip_grad_by_norm: 35.
  optimizer:
    momentum: 0.9
    type: Momentum
  regularizer:
    factor: 0.0005
    type: L2


YOLOv3:
  backbone: ResNet
  neck: PPYOLOPAN
  yolo_head: YOLOv3Head
  post_process: BBoxPostProcess

ResNet:
  depth: 50
  variant: d
  return_idx: [1, 2, 3]
  dcn_v2_stages: [3]
  freeze_at: -1
  freeze_norm: false
  norm_decay: 0.

PPYOLOPAN:
  drop_block: true
  block_size: 3
  keep_prob: 0.9
  spp: true

YOLOv3Head:
  anchors: [[10, 13], [16, 30], [33, 23],
            [30, 61], [62, 45], [59, 119],
            [116, 90], [156, 198], [373, 326]]
  anchor_masks: [[6, 7, 8], [3, 4, 5], [0, 1, 2]]
  loss: YOLOv3Loss
  iou_aware: true
  iou_aware_factor: 0.4

YOLOv3Loss:
  ignore_thresh: 0.7
  downsample: [32, 16, 8]
  label_smooth: false
  scale_x_y: 1.05
  iou_loss: IouLoss
  iou_aware_loss: IouAwareLoss

IouLoss:
  loss_weight: 2.5
  loss_square: true

IouAwareLoss:
  loss_weight: 1.0

BBoxPostProcess:
  decode:
    name: YOLOBox
    conf_thresh: 0.01
    downsample_ratio: 32
    clip_bbox: true
    scale_x_y: 1.05
  nms:
    name: MatrixNMS
    keep_top_k: 100
    score_threshold: 0.01
    post_threshold: 0.01
    nms_top_k: -1
    background_label: -1