Towards 3D Object Detection with 2D Supervision
This reduces annotation costs for 3D perception tasks, making it more accessible for applications like autonomous driving, though it is incremental as it builds on existing 2D-to-3D methods.
The paper tackles the high cost of 3D annotations for object detection by introducing a hybrid training framework that uses 2D supervision to train a 3D detector, achieving nearly 90% of fully-supervised performance on nuScenes with only 25% 3D annotations.
The great progress of 3D object detectors relies on large-scale data and 3D annotations. The annotation cost for 3D bounding boxes is extremely expensive while the 2D ones are easier and cheaper to collect. In this paper, we introduce a hybrid training framework, enabling us to learn a visual 3D object detector with massive 2D (pseudo) labels, even without 3D annotations. To break through the information bottleneck of 2D clues, we explore a new perspective: Temporal 2D Supervision. We propose a temporal 2D transformation to bridge the 3D predictions with temporal 2D labels. Two steps, including homography wraping and 2D box deduction, are taken to transform the 3D predictions into 2D ones for supervision. Experiments conducted on the nuScenes dataset show strong results (nearly 90% of its fully-supervised performance) with only 25% 3D annotations. We hope our findings can provide new insights for using a large number of 2D annotations for 3D perception.