2021/03/02 by Benjamin Caine, Rebecca Roelofs, Caine, Benjamin +11 · 5 citations
Computer Science · Engineering · #Adaptation (eye) #Advanced Neural Network Applications #Artificial intelligence #Bounding overwatch #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Data mining #Domain (mathematical analysis) #Domain Adaptation and Few-Shot Learning #Domain adaptation #Exploit #FOS: Computer and information sciences #Image (mathematics) #Labeled data #Machine Learning (cs.LG) #Machine learning #Minimum bounding box #Object (grammar) #Object detection #Pattern recognition (psychology) #Robotics and Sensor-Based Localization #Scalability #Set (abstract data type) #Software deployment #Training set #cs.CV #cs.LG
paper · pdf · doi:10.48550/arxiv.2103.02093
published in arXiv (Cornell University) (Cornell University)
arxiv created 2021/03/02 · openalex publication_date 2021/03/02 · arxiv updated 2021/03/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06
To safely deploy autonomous vehicles, onboard perception systems must work reliably at high accuracy across a diverse set of environments and geographies. One of the most common techniques to improve the efficacy of such systems in new domains involves collecting large labeled datasets, but such datasets can be extremely costly to obtain, especially if each new deployment geography requires additional data with expensive 3D bounding box annotations. We demonstrate that pseudo-labeling for 3D object detection is an effective way to exploit less expensive and more widely available unlabeled data, and can lead to performance gains across various architectures, data augmentation strategies, and sizes of the labeled dataset. Overall, we show that better teacher models lead to better student models, and that we can distill expensive teachers into efficient, simple students. Specifically, we demonstrate that pseudo-label-trained student models can outperform supervised models trained on 3-10 times the amount of labeled examples. Using PointPillars [24], a two-year-old architecture, as our student model, we are able to achieve state of the art accuracy simply by leveraging large quantities of pseudo-labeled data. Lastly, we show that these student models generalize better than supervised models to a new domain in which we only have unlabeled data, making pseudo-label training an effective form of unsupervised domain adaptation.