Back to the catalogue
Image classificationFeaturedManual review
Classify Autonomous Vehicle Sensor Annotations & Edge Cases
Review high-resolution camera frames and verify 3D bounding boxes around pedestrians, cyclists, and lane barriers.
18 min estimated240 DoPoints rewardFocused difficultyDesktop38 of 150 slots left
About this task
An autonomy team is cleaning a sensor dataset before it trains a perception model. You review camera frames with overlaid bounding boxes and confirm, correct or reject each annotation.
What is expected
- A verdict for every frame in the batch
- Corrected labels where a box is wrong or missing
- Edge cases flagged with a short explanation
Important rules
- Work at 100% zoom; do not resize the frame window.
- Reject frames where the sensor data is corrupted.
- Every correction needs a one-line reason.
Requirements
- Desktop required for the annotation grid
- Around 18 minutes for the full batch
How this is validated
A reviewer reads your submission and compares it against the brief. Most reviews finish within 24 hours, and you are notified either way.
Before you accept
The deadline starts when you accept, not when you open the task. If you cannot finish, abandon it from the task screen rather than letting it expire. Abandoning carries the smaller penalty and keeps more of your approval rate.