Import and export

Import an existing dataset

You can create a project from a folder or zip that already contains images (and optionally YOLO labels).

Expected layouts

YOLO-style (preferred):

dataset/
  images/
    train/   # or flat images/
    ...
  labels/
    train/
    ...
  data.yaml   # optional but recommended

Also recognized: classes.txt, obj.names, and plain image folders (images only — no labels).

UI

On New project, use the import tab:

Pick a local folder path or upload a zip
Optionally use source as location (folder import only) to keep images in place instead of copying into ~/.oneopen/projects/...

API

Multipart (folder path and/or zip file):

POST /api/projects/import

JSON folder import:

POST /api/projects/import/folder
Content-Type: application/json

{
  "name": "My YOLO set",
  "folder_path": "D:/datasets/traffic",
  "use_source_as_location": true,
  "project_type": "object_detection"
}

Imported boxes/polygons are stored with source=import. Class names come from data.yaml / names files when present.

Assign splits

Before generating a version:

POST /api/projects/{project_id}/splits?train=70&valid=20&test=10

Images are randomly assigned to train / valid / test according to the percentages.

Export

UI: Project → Export panel → choose format and version.

YOLO (zip)

Creates a zip under ~/.oneopen/exports/ with images, labels, data.yaml, and pipeline metadata. Preprocess/augment from the selected version are applied when building the export.

POST /api/projects/{project_id}/export
{"format": "yolo", "version_id": 1, "include_unannotated": false}

Download:

GET /api/exports/{filename}

COCO (JSON)

Writes a COCO-style JSON file (annotations + image metadata). Image binary files are not packaged in the COCO export.

POST /api/projects/{project_id}/export
{"format": "coco", "version_id": 1, "include_unannotated": false}

Warning

voc may appear in schema comments / discussions but Pascal VOC export is not implemented. Non-coco formats currently fall through to the YOLO exporter.