High-Precision Video Annotation for Autonomous Navigation (Atlas Capture)
Internal Proprietary ToolingVideoPolygonObject Detection
Performed frame-by-frame video annotation on the proprietary Atlas Capture platform to support the development of computer vision models. My work focused on identifying and tracking dynamic objects (pedestrians, vehicles, cyclists) across temporal sequences.
Key responsibilities included:
Applying Polygon masks and Polyline annotations with sub-pixel accuracy to define road boundaries and obstacles.
Utilizing Object Tracking techniques to ensure ID consistency for entities moving through occlusions and variable lighting.
Adhering to strict "Ground Truth" guidelines to maintain a 98%+ quality audit score.
Identifying and reporting dataset edge cases, such as rare weather conditions or unconventional vehicle types.
Performed frame-by-frame video annotation on the proprietary Atlas Capture platform to support the development of computer vision models. My work focused on identifying and tracking dynamic objects (pedestrians, vehicles, cyclists) across temporal sequences. Key responsibilities included: Applying Polygon masks and Polyline annotations with sub-pixel accuracy to define road boundaries and obstacles. Utilizing Object Tracking techniques to ensure ID consistency for entities moving through occlusions and variable lighting. Adhering to strict "Ground Truth" guidelines to maintain a 98%+ quality audit score. Identifying and reporting dataset edge cases, such as rare weather conditions or unconventional vehicle types.
2024