Setting multiple directories for training and validation in yolo_v4_train_resnet18.txt

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi I tarin with TAO yolov4
have 3 directories for training and 3 for validation (all set in the docker via .tao_mounts.json)
Now I have noticed that if I set them in the yolo_v4_train_resnet18.txt in the following way I get an error regarding the validation folders only:

dataset_config {
  data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset1/training/label_2"
      image_directory_path: "/workspace/tao-experiments/data/dataset1/training/image_2"
  }
    data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset2/training/label_2"
      image_directory_path: "/workspace/tao-experiments/data/dataset2/training/image_2"
  }
    data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset3/training/label_2"
      image_directory_path: "/workspace/tao-experiments/data/dataset3/training/image_2"
  }
  
  


  include_difficult_in_training: true
  image_extension: "png"
  
  target_class_mapping {
      key: "fff"
      value: "fff"
  }
  target_class_mapping {
      key: "eee"
      value: "eee"
  }
  target_class_mapping {
      key: "ddd"
      value: "ddd"
  }
  target_class_mapping {
      key: "ccc"
      value: "ccc"
  }
  target_class_mapping {
      key: "bbb"
      value: "bbb"
  }
  target_class_mapping {
      key: "aaa"
      value: "aaa"
  }


validation_data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset1/val/label"
      image_directory_path: "/workspace/tao-experiments/data/dataset1/val/image"
      }
      
validation_data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset2/val/label"
      image_directory_path: "/workspace/tao-experiments/data/dataset2/val/image"
      }
      
validation_data_sources: {
      label_directory_path: "/workspace/tao-experiments/data/dataset3/val/label"
      image_directory_path: "/workspace/tao-experiments/data/dataset3/val/image"
      }

The only thing helped was to change the validation part like this:

validation_data_sources: [
  {
    label_directory_path: "/workspace/tao-experiments/data/dataset1/val/label"
    image_directory_path: "/workspace/tao-experiments/data/dataset1/val/image"
  },
  {
    label_directory_path: "/workspace/tao-experiments/data/dataset2/val/label"
    image_directory_path: "/workspace/tao-experiments/data/dataset2/val/image"
  },
  {
    label_directory_path: "/workspace/tao-experiments/data/dataset3/val/label"
    image_directory_path: "/workspace/tao-experiments/data/dataset3/val/image"

  }
]
  1. why this happens?
    2)is it ok to format also the training part in the same way as validation for more consistency?
3

Please refer to YOLOv4

image
image2352×247 30.6 KB

4

Hi please answer more speciffaclly to what I am asking.

5

Can you share the error log when you set as above? According to the user guide, seems that it is correct.

6

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

Hi,
I can run with below format.

  validation_data_sources: {
    image_directory_path: "/workspace/tlt-experiments/data/kitti_dataset/val/image"
    label_directory_path: "/workspace/tlt-experiments/data/kitti_dataset/val/label"
  }
  validation_data_sources: {
    image_directory_path: "/workspace/tlt-experiments/data/kitti_dataset/training/image_2"
    label_directory_path: "/workspace/tlt-experiments/data/kitti_dataset/training/label_2"
  }
8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.