Sample datasets

Here are some small datasets you can use to familiarize yourself with Latent Assisted Label.

Clicking the link will download the dataset. To load a dataset for the first time, use the Create Dataset from Directory operator.

Create Dataset from Directory

Give the dataset a name, provide a relative or absolute path to the dataset directory, and click Execute.

Once you've loaded a dataset, it will always be available to load in the Select Dataset menu.

Load Dataset

• Sonar

  Side-scan sonar images for Find Everything demos. Recommended settings: max_area = 0.05, min_area = 0.001. Derived from the Side-scan sonar imaging for mine detection dataset: 1,170 images taken between 2010 and 2021 using Teledyne Marine Gavia Autonomous Underwater Vehicle (AUV). Supports object detection, classification, and segmentation for mine-like and non-mine-like underwater objects.

• Overhead

  Overhead aerial images for Find Everything and Find Description demos. Recommended settings: max_area = 0.05, min_area = .001, nms_iou_threshold = 0.3, sam_nms_threshold = 0.3. Sourced from the Virtual Aerial Image Dataset (VALID).

• Airplanes, people, and boats

  Images with airplanes, people, and boats for Find Description demos. Sourced from Open Images V7, this dataset demonstrates category diversity and cross-domain visual grounding.

• Cars and people

  Images with cars and people for Find Description demos. These images, handpicked from the Open Images V7 dataset, are ideal for testing urban scene understanding and multi-object detection.

Sample video

• Walking

  A video of people walking in a city. This video is sourced from Pexels and is suitable for testing the Follow Object operation.