Bootstrapping Autonomous Driving Radars with Self-Supervised Learning
This work addresses the cost and difficulty of annotating radar data for autonomous vehicles, offering an incremental improvement in perception accuracy.
The paper tackles the problem of training radar models for autonomous driving by proposing a self-supervised learning framework that uses unlabeled radar data paired with camera images to pre-train embeddings, resulting in a 5.8% improvement in mAP for object detection compared to supervised baselines.
The perception of autonomous vehicles using radars has attracted increased research interest due its ability to operate in fog and bad weather. However, training radar models is hindered by the cost and difficulty of annotating large-scale radar data. To overcome this bottleneck, we propose a self-supervised learning framework to leverage the large amount of unlabeled radar data to pre-train radar-only embeddings for self-driving perception tasks. The proposed method combines radar-to-radar and radar-to-vision contrastive losses to learn a general representation from unlabeled radar heatmaps paired with their corresponding camera images. When used for downstream object detection, we demonstrate that the proposed self-supervision framework can improve the accuracy of state-of-the-art supervised baselines by $5.8\%$ in mAP. Code is available at \url{https://github.com/yiduohao/Radical}.