The deployment of 3D detectors strikes one of the major challenges in real-world self-driving scenarios.
However, the typical convolution ignores the radial symmetry of the BEV features and increases the difficulty of the detector optimization.
The goal of this work is to establish a scalable pipeline for expanding an object detector towards novel/unseen categories, using zero manual annotations.
One-stage object detection is commonly implemented by optimizing two sub-tasks: object classification and localization, using heads with two parallel branches, which might lead to a certain level of spatial misalignment in predictions between the two tasks.
Ranked #172 on Object Detection on COCO test-dev
Specifically, EBL increases the intensity of the adjustment of the decision boundary for the weak classes by a designed score-guided loss margin between any two classes.
Ranked #8 on Object Detection on LVIS v1.0 val
Data collection and annotation are time-consuming in machine learning, expecially for large scale problem.