Object Track
Object Track allows users to visualize the entire track of a selected cuboid across a sequence. When enabled, it displays all instances of that cuboid - before and after the current frame - providing a full view of the object’s trajectory. This feature is particularly useful for reviewing dynamic objects and catching inconsistencies in their annotations, such as misalignment, missed frames, or sudden shifts. Object Track is a good and fast review tool that enhances annotation quality while minimizing the time spent scrubbing through frames.
Visualize and verify an object’s full trajectory across frames to ensure consistent, high-quality tracking.
Use Cases
Autonomous Driving – Moving Vehicles and Pedestrians Visually confirm consistent tracking of dynamic objects like cars or pedestrians over time to ensure high-quality training data.
Robotics – Tracking People or Equipment Indoors Use Object Track to validate that moving people or tools in indoor robotics datasets are annotated consistently across frames.
Agri-Tech – Monitoring Moving Machinery or Livestock Confirm that tractors, drones, or animals are correctly tracked over time in long sequences with intermittent motion or occlusions.
Benefits
Good: Increases annotation quality by making it easy to spot tracking gaps, misaligned cuboids, or frame skips.
Fast: Saves time during review by surfacing the full track in one view—eliminating the need to click through each frame.
Lean: Lightweight toggle-based feature that fits seamlessly into existing workflows with no added complexity.
Steps to Use Object Track
Last updated