MaskBEV: Joint Object Detection and Footprint Completion for Bird's-eye View 3D Point Clouds
This addresses limitations in LiDAR-based object detection for autonomous driving by simplifying the detection process, but it appears incremental as it builds on existing mask-based approaches in a new context.
The paper tackles object detection in LiDAR point clouds by proposing MaskBEV, a bird's-eye view mask-based detector that predicts instance masks for object footprints, eliminating the need for bounding box regression. It achieves evaluation on SemanticKITTI and KITTI datasets, though no concrete performance numbers are provided in the abstract.
Recent works in object detection in LiDAR point clouds mostly focus on predicting bounding boxes around objects. This prediction is commonly achieved using anchor-based or anchor-free detectors that predict bounding boxes, requiring significant explicit prior knowledge about the objects to work properly. To remedy these limitations, we propose MaskBEV, a bird's-eye view (BEV) mask-based object detector neural architecture. MaskBEV predicts a set of BEV instance masks that represent the footprints of detected objects. Moreover, our approach allows object detection and footprint completion in a single pass. MaskBEV also reformulates the detection problem purely in terms of classification, doing away with regression usually done to predict bounding boxes. We evaluate the performance of MaskBEV on both SemanticKITTI and KITTI datasets while analyzing the architecture advantages and limitations.