Quickstart

Import a local image directory, configure its annotation schema, and inspect the dataset in the Explorer.

Prerequisites

  • Pixano running (see Installation).
  • A flat directory containing only JPEG or PNG files (no subdirectories).
Note

The example folder, my_images_dataset, contains three images. Record counts depend on the selected files.

Import with the wizard

  1. Open Pixano in a browser. In the empty Library, click Import dataset to open the import wizard.
The empty Library: 'Your library is empty' with the Import dataset button.
Empty Library.
  1. In Setup, use the Source pane on the left to select Raw media.
  2. In Annotation setup on the right, select Image annotation. Bounding boxes and Masks are enabled by default; Polygons & lines is optional. Add Object attributes such as category (String) if required by the annotation schema.
  3. In Source, click Choose a folder and select my_images_dataset. After upload, Detected layout reports 3 records, each with one image view.
  4. Leave Dataset name empty to use the folder name. Keep Overwrite existing dataset unchecked for a new dataset, then click Review import.
Setup with Source on the left and Annotation setup on the right: my_images_dataset is uploaded, 3 records are detected, and bounding boxes and masks are selected.
Setup: source configuration and annotation schema.
  1. In Review, check the record count, Media views, Annotation setup, and Sample records. Analysis does not create the dataset. Use Edit setup to make changes, or click Start import.
Review import showing 3 records, the image view, bounding box and mask tools, and sample records.
Review the import plan before starting the job.
  1. The wizard closes and the import runs in the background. The job tray shows progress and a Cancel control. On completion, the Library updates automatically with the new dataset.
The Library with the new my_images_dataset card and an Imported status in the background job tray.
The Library updates automatically when the background import completes.
Note

The same Setup → Review flow supports video, Visual Q&A, and image–text annotation tasks, plus existing Pixano, MS COCO, and LeRobot datasets. Select the data format in Source and configure applicable tools and attributes in Annotation setup. In a raw media directory, train/, val/, and test/ subdirectories become splits.

Inspect the dataset

Click the dataset card. The Explorer lists the records in a gallery, one per image. Click a record to open it in the annotation Workspace.

The Explorer showing the three records of my_images_dataset in the gallery.
Explorer, gallery view.

Each image is stored as a record with a single view. Annotated objects are entities; each bounding box, mask, or polygon attached to an entity is an annotation. Review summarizes the views, annotation tools, and custom attributes. See Key Concepts for the data model.

Esc
Type to search the documentation