Scanabull has created a machine learning algorithm to determine the weight of a cow using 3D imagery. The data model requires very accurate data in order to improve its predictions. This is where your efforts in data labelling become so useful.
- Login: Access the worker portal using the credentials emailed to you and create a new password
- Quality over Speed: Thoroughness is more important than speed. Accuracy in labeling is crucial for this task.
Your objective is to identify and mark all points in the 3D point cloud that are NOT part of a cow. Use the paint brush tool to highlight these non-cow points.
When you first enter the 3D view, you may need to enlarge the points to clearly see the cow structure. Increase the point size until you can make out the image of a cow.
Use the paint brush tool to mark all points that are not part of the cow:
Technical Tips:
- Use zoom controls - zoom in/out to make sure all points that should be selected are properly marked
- Switch modes easily - when using the brush, press the "M" button to return to rotation mode for the point cloud
After marking some points, rotate the 3D model and inspect the cow from different angles to ensure you've captured all non-cow points. This multi-angle approach is essential for complete accuracy.
- Rotate and inspect from multiple angles to ensure comprehensive coverage
- Error on the side of labelling not-cow When in doubt, its better to mark a point as being not-cow
- Take your time - accuracy is more important than speed
- Mark all non-cow points using the paint brush tool
- Don't forget distant points - be sure to mark non-cow points that are far away from the cow
- Pay special attention to legs - points close to the legs can be tricky, so take your time there
- Double-check your work by rotating the model to view from different perspectives