参数记录

发布时间 2023-04-21 10:41:26作者: yonuyeung

dog_detection.yml路径一键修改:

metric: COCO
num_classes: 4

TrainDataset:
  !COCODataSet
    image_dir: images
    anno_path: annotations/train.json
    dataset_dir: /home/aistudio/data/robot_dog_dataset/detection/
    data_fields: ['image', 'gt_bbox', 'gt_class', 'is_crowd']

EvalDataset:
  !COCODataSet
    image_dir: images
    anno_path: annotations/val.json
    dataset_dir: /home/aistudio/data/robot_dog_dataset/detection/

TestDataset:
  !ImageFolder
    anno_path: annotations/val.json # also support txt (like VOC's label_list.txt)
    dataset_dir: /home/aistudio/data/robot_dog_dataset/detection/ # if set, anno_path will be 'dataset_dir/anno_path'
View Code

2023年4月21日10:39

环境:V100 32G

dog_detection.yml

num_classes:4

optimizer_80e.yml

epoch: 600
base_lr: 0.000125

ppyoloe_plus_reader.yml

TrainReader:
  batch_transforms:
    - NormalizeImage: {mean: [0.9335411,0.9335429,0.9335709], std: [0.09545881,0.09532484,0.0950378], norm_type: none}
  batch_size: 32
 
EvalReader:
  sample_transforms:
    - NormalizeImage: {mean: [0.9335411,0.9335429,0.9335709], std: [0.09545881,0.09532484,0.0950378], norm_type: none}
  batch_size: 4
 
TestReader:
  sample_transforms:
    - NormalizeImage: {mean: [0.9335411,0.9335429,0.9335709], std: [0.09545881,0.09532484,0.0950378], norm_type: none}
 

yolov3_darknet53.yml

具体路径:/home/aistudio/PaddleDetection/configs/yolov3/_base_/yolov3_darknet53.yml
YOLOv3Loss:
  label_smooth: true