PatchContrast: Self-Supervised Pre-training for 3D Object Detection
This addresses the data annotation bottleneck for autonomous vehicles, though it appears incremental as it builds on existing self-supervised pre-training approaches.
The paper tackles the problem of expensive annotated data for 3D object detection by introducing PatchContrast, a self-supervised pre-training framework that uses proposal-level and patch-level abstraction to learn representations from unlabeled point clouds. It shows that this method outperforms state-of-the-art models on three common 3D detection datasets.
Accurately detecting objects in the environment is a key challenge for autonomous vehicles. However, obtaining annotated data for detection is expensive and time-consuming. We introduce PatchContrast, a novel self-supervised point cloud pre-training framework for 3D object detection. We propose to utilize two levels of abstraction to learn discriminative representation from unlabeled data: proposal-level and patch-level. The proposal-level aims at localizing objects in relation to their surroundings, whereas the patch-level adds information about the internal connections between the object's components, hence distinguishing between different objects based on their individual components. We demonstrate how these levels can be integrated into self-supervised pre-training for various backbones to enhance the downstream 3D detection task. We show that our method outperforms existing state-of-the-art models on three commonly-used 3D detection datasets.