Artificial Dummies for Urban Dataset Augmentation
This work is significant for the autonomous driving industry, which requires extremely high detection accuracy in rare and challenging scenarios, by providing a method to generate diverse and controlled training data.
The paper addresses the scarcity of diverse pedestrian data for training detectors by proposing DummyNet, a data generator that synthesizes urban scenes with people in arbitrary poses and appearances, embedded in various backgrounds. This augmentation method significantly improves the performance of existing person detectors, achieving a 17% reduction in log-average miss rate for night-time detection when trained only with day-time data.
Existing datasets for training pedestrian detectors in images suffer from limited appearance and pose variation. The most challenging scenarios are rarely included because they are too difficult to capture due to safety reasons, or they are very unlikely to happen. The strict safety requirements in assisted and autonomous driving applications call for an extra high detection accuracy also in these rare situations. Having the ability to generate people images in arbitrary poses, with arbitrary appearances and embedded in different background scenes with varying illumination and weather conditions, is a crucial component for the development and testing of such applications. The contributions of this paper are three-fold. First, we describe an augmentation method for controlled synthesis of urban scenes containing people, thus producing rare or never-seen situations. This is achieved with a data generator (called DummyNet) with disentangled control of the pose, the appearance, and the target background scene. Second, the proposed generator relies on novel network architecture and associated loss that takes into account the segmentation of the foreground person and its composition into the background scene. Finally, we demonstrate that the data generated by our DummyNet improve performance of several existing person detectors across various datasets as well as in challenging situations, such as night-time conditions, where only a limited amount of training data is available. In the setup with only day-time data available, we improve the night-time detector by $17\%$ log-average miss rate over the detector trained with the day-time data only.