Our Achivements
Project Goals
- Development of pre-labeler for 2D & 3D annotation tools
- Customizing of open source annotation tools
Accomplishments
Automatic 2D Image Bounding Box and Polygon Labelling
- Support a large and growing number of labels for 2D labeling : cars, pedestrians, traffic light, cyclists, traffic sign, construction cones to free-space, lane markings, road curbs, road boundaries.
- Integrated with state-of-the-art deep learning algorithms for auto annotation
Automatic 3D Lidar Object Detection and Tracking / Sensor Fusion Cuboid Segmentation
- Support labels for 3D : vehicles, cyclists, pedestrians, wheelchairs, strollers, traffic cones and barriers.
- 3D object tracking for temporally linked frames
Annotation Specifications
- Writing technical documents for annotation policy for detection : Object, Lane, Free-space, Traffic sing & light, General, Vehicle, etc
Customizing open source annotation tools
- Integration & Customization of open source tools according to customer’s req.
- Support customized LIDAR Tools and our own image pre-labeler with task mgmt.
Technologies Used
- Django, Bootstrap studio, Pytorch, Docker, etc