Tiny-YOLO object detection supplemented with geometrical data
This work addresses the need for more accurate object detection in autonomous robots, though it is incremental as it builds upon the existing Tiny-YOLO framework with a specific geometric enhancement.
The authors tackled the problem of improving object detection precision for autonomous robots by incorporating prior knowledge about scene geometry, specifically assuming a planar scene, and achieved a higher mAP with minimal computational overhead compared to standard RGB-based detection.
We propose a method of improving detection precision (mAP) with the help of the prior knowledge about the scene geometry: we assume the scene to be a plane with objects placed on it. We focus our attention on autonomous robots, so given the robot's dimensions and the inclination angles of the camera, it is possible to predict the spatial scale for each pixel of the input frame. With slightly modified YOLOv3-tiny we demonstrate that the detection supplemented by the scale channel, further referred as S, outperforms standard RGB-based detection with small computational overhead.