BYOD - COCO format¶
Category: data_generator.vision.detection.2d
Version: 1.1.0
UUID: edd4d9cf30659132f9eaf64f044fc13c14ff410bdd097b5f6ab8044e8d691696
Execution context: leip_af
Choice Priority: 10.0
Subcomponent Parameters¶
| Name | Synonyms | Allowable Categories |
|---|---|---|
| Training Augmentation | data.augmentation.training |
augmentation.vision.2d.training |
| Validation Augmentation | data.augmentation.validation |
augmentation.vision.2d.validation |
Value Parameters¶
| Name | Synonyms | Type | Values |
|---|---|---|---|
| Dataset name | data_generator.dataset_name |
scalar | string |
| Class label indexing scheme | data_generator.label_indexing |
choice | 0-indexed-no-background (COCO index starts with 0, no background class)1-indexed-no-background (COCO index starts with 1, no background class)0-indexed-with-background (COCO index starts with 0, 0 is background class) |
| Root Path | data_generator.root_path |
scalar | path |
| Training annotation JSON file | data_generator.train_annotations_json |
scalar | path |
| Training image subdir | data_generator.train_images_dir |
scalar | path |
| Validation annotation JSON file | data_generator.val_annotations_json |
scalar | path |
| Validation image subdir | data_generator.val_images_dir |
scalar | path |
| Number of classes | data.number_of_classes |
scalar | int |
Constraints¶
- Number of classes has to be positive (meta.nclasses > 0)
- Need a root path for dataset (len(str(actual.root_path)) > 0)