Validate your dataset structure to use NetsPresso

How to use

1. Create python virtual environment and activate it

python3 -m venv .venv && source .venv/bin/activate

2. Install requirements

pip3 install -r requirements.txt

3. Run CLI

Run the command below to validate when the data is ready-to-use. If you do not get any error message, you are all set! If error occurs, please refer to the error message to resolve the issue.

🔹When uploading via Netspresso web page

python3 run.py --format {dataset_format} \
               --task {task type} \
               --storage_type {storage_type} \
               --yaml_path {data.yaml file path(only required in yolo format)} \
               --train_dir {train foler path} --valid_dir {validation folder path} \
               --test_dir {test folder path} \
               --output_dir {output path} \
               --id2label {id2label.json file path(only required in unet format)}

Argument	Values	Description	Required
--format	yolo / voc / coco / unet / imagenet	format of the dataset	O
--task	image_classification / object_detection / semantic_segmentation	task of the dataset	O
--storage_type	s3 / local	storage type of the dataset	O
--yaml_path	ex) ./coco_dataset/data.yaml	yaml file path for yolo format dataset	only when using yolo format
--train_dir	ex) ./coco_dataset/train	train dataset folder path	O
--test_dir	ex) ./coco_dataset/test	test dataset folder path	optional
--valid_dir	ex) ./coco_dataset/val	valid dataset folder path	optional
--output_dir	ex) ./validation_dataset	output folder where validation results will be stored	optional
--id2label_path	ex) ./unet_dataset/id2label.json	id2label file path for unet format dataset	only when using unet format

🔹When using dataset in personal server

python3 run.py --format {dataset_format} \
               --task {task type} \
               --storage_type {storage_type} \
               --dataset_root_path {dataset root folder path} \
               --yaml_path {data.yaml file path(only required in yolo format)} \
               --id2label {id2label.json file path(only required in unet format)} \
               --server_info_path {server_info_netspresso.json file path}

arguments:

Argument	Values	Description	Required
--format	yolo / voc / coco / unet / imagenet	format of the dataset	O
--task	image_classification / object_detection / semantic_segmentation	task of the dataset	O
--storage_type	s3 / local	storage type of the dataset	O
--dataset_root_path	ex) ./coco_dataset/	dataset root path	O
--yaml_path	ex) ./coco_dataset/data.yaml	yaml file path for yolo format dataset	only when using yolo format
--id2label_path	ex) ./unet_dataset/id2label.json	id2label file path for unet format dataset	only when using unet format
--server_info_path	ex) ./netspresso/server_info_netspresso.json	A server information file which is generated upon registering a personal training server	O

Validation success case

certification.np contains certification and file information which created together. Make sure that the files are not changed. If any file changes, the certification will not work and may not enable you to upload dataset to NetsPresso.

netspresso@netspresso:~/NetsPresso-ModelSearch-Dataset-Validator$ python3 run.py --format yolo --yaml_path ./data.yaml --train_dir ./dataset/train --valid_dir ./dataset/val --test_dir ./dataset/test --output_dir ./output
Start dataset validation.
Validation completed! Now try your dataset on NetsPresso!

🔹When uploading via Netspresso web page

Five(or Four) files will be created at the selected output path.

train.zip
val.zip
test.zip
certification.np
data.yaml

🔹When using dataset in personal server

Two files will be created as shown below at the dataset root path.

certification.np
data.yaml

Validation fail case

For more detail, please see [Validation check list][validationchecklist]

In case of validation fail with traceback, please read exception error message.

netspresso@netspresso:~/NetsPresso-ModelSearch-Dataset-Validator$ python3 run.py --format yolo --task object_detection --yaml_path ./data.yaml --train_dir ./dataset/train --valid_dir ./dataset/val --test_dir ./dataset/test --output_dir ./output
Start dataset validation.
Traceback (most recent call last):
  File "run.py", line 13, in <module>
    validate(dir_path, num_classes, dataset_type)
  File "/hdd1/home/NetsPresso-ModelSearch-Dataset-Validator/src/utils.py", line 307, in validate
    yaml_label, errors = validate_data_yaml(dir_path, num_classes, errors)
  File "/hdd1/home/NetsPresso-ModelSearch-Dataset-Validator/src/utils.py", line 199, in validate_data_yaml
    raise YamlException("There is no 'names' in data.yaml.")
src.exceptions.YamlException: There is no 'names' in data.yaml.

In case of validation fail with Validation error, please check 'validation_result.txt'., please check validation_result.txt file to resolve failure.

netspresso@netspresso:~/NetsPresso-ModelSearch-Dataset-Validator$ python3 run.py --format yolo --task object_detection --yaml_path ./data.yaml --train_dir ./dataset/train --valid_dir ./dataset/val --test_dir ./dataset/test --output_dir ./output
Start dataset validation.
Validation error, please check 'validation_result.txt'.

And contents of 'validation_result.txt are like below.

There is no image file for annotation file 'yolo/train/labels/000000000025.txt'
There is no image file for annotation file 'yolo/test/labels/000000000337.txt'

Dataset Structure

Please read this link: Dataset Structure