Semi-supervised lane detection with Deep Hough Transform
This work addresses the need for cheaper and more scalable lane detection systems for autonomous driving, though it is incremental as it builds on existing semi-supervised methods.
The paper tackles the problem of reducing dependency on manual annotations for lane detection by leveraging unlabeled data, proposing a novel loss function based on the Deep Hough Transform that improves performance on CULane and TuSimple datasets.
Current work on lane detection relies on large manually annotated datasets. We reduce the dependency on annotations by leveraging massive cheaply available unlabelled data. We propose a novel loss function exploiting geometric knowledge of lanes in Hough space, where a lane can be identified as a local maximum. By splitting lanes into separate channels, we can localize each lane via simple global max-pooling. The location of the maximum encodes the layout of a lane, while the intensity indicates the the probability of a lane being present. Maximizing the log-probability of the maximal bins helps neural networks find lanes without labels. On the CULane and TuSimple datasets, we show that the proposed Hough Transform loss improves performance significantly by learning from large amounts of unlabelled images.