dataset_augment

Create a local augmented YOLO dataset copy using deterministic transforms.

Overview

dataset_augment applies image augmentation techniques to an existing YOLO dataset and writes the augmented images and labels to a new output directory. The original dataset is never modified.

Input Schema

Parameter	Type	Required	Default	Description
dataset_path	string	Yes	—	Path to the source YOLO dataset directory.
techniques	list[string]	No	["flip_h"]	Augmentation techniques to apply: flip_h, flip_v, cutout.
multiplier	integer	No	3	Number of augmented copies per source image.
output_path	string	Yes	—	Path to write the augmented dataset.

Output Schema

Field	Type	Description
dataset_path	string	Resolved source dataset path.
output_path	string	Resolved output dataset path.
source_images	integer	Number of source images found.
generated_images	integer	Total augmented images created.
techniques	list[string]	Techniques applied.
manifest_path	string	Path to the augmentation manifest JSON.

Examples

CLI

fovux-mcp serve --http
curl -X POST http://127.0.0.1:7823/tools/dataset_augment \
  -H "Authorization: Bearer $(cat ~/.fovux/auth.token)" \
  -H "Content-Type: application/json" \
  -d '{"dataset_path": "/data/yolo_set", "output_path": "/data/yolo_aug", "techniques": ["flip_h", "flip_v"], "multiplier": 2}'

Python

from fovux.tools.dataset_augment import dataset_augment
result = dataset_augment("/data/yolo_set", techniques=["flip_h"], multiplier=3, output_path="/data/yolo_aug")

Notes & Limits

• The data.yaml file is copied from the source to the output directory.
• Labels are adjusted to match the applied augmentation (e.g., bounding box x-coordinates are mirrored for flip_h).
• cutout places a black rectangle over the center 30% of the image.
• Large datasets with high multipliers may consume significant disk space.

Failure Modes

• FovuxDatasetNotFoundError if the source dataset path does not exist.
• ValueError if output_path is not specified.
• Validation error if data.yaml is missing or malformed.