Lidar-Camera Co-Training for Semi-Supervised Road Detection
This work addresses the cost and effort of data annotation for road detection systems, which is an incremental improvement in semi-supervised learning for autonomous driving.
The paper tackles the problem of expensive and time-consuming manual labeling for road detection by applying a semi-supervised co-training approach, resulting in F1-score improvements of 1.12 to 8.14 percentage points for camera- and lidar-based detectors and achieving high performance on the KITTI benchmark with only 36 labeled examples.
Recent advances in the field of machine learning and computer vision have enabled the development of fast and accurate road detectors. Commonly such systems are trained within a supervised learning paradigm where both an input sensor's data and the corresponding ground truth label must be provided. The task of generating labels is commonly carried out by human annotators and it is notoriously time consuming and expensive. In this work, it is shown that a semi-supervised approach known as co-training can provide significant F1-score average improvements compared to supervised learning. In co-training, two classifiers acting on different views of the data cooperatively improve each other's performance by leveraging unlabeled examples. Depending on the amount of labeled data used, the improvements ranged from 1.12 to 6.10 percentage points for a camera-based road detector and from 1.04 to 8.14 percentage points for a lidar-based road detector. Lastly, the co-training algorithm is validated on the KITTI road benchmark, achieving high performance using only 36 labeled training examples together with several thousands unlabeled ones.