The detection of road agents, such as vehicles and pedestrians are central in autonomous driving. Self-Supervised Learning (SSL) has been proven to be an effective technique for learning discriminative feature representations for image classification, alleviating the need for labels, a remarkable advancement considering how timeconsuming and expensive labeling can be in autonomous driving. In this paper, we investigate the effectiveness of contrastive SSL techniques such as BYOL and MOCO on the object (agent) detection task using the ROad event Awareness Dataset (ROAD) and BDD100K benchmarks. Our experiments show that using self-supervised pretraining, we can achieve a 3.96 and 0.78 percentage points improvement on the AP50 metric on the ROAD and BDD100K benchmarks for the object detection task compared to supervised pretraining. Extensive comparisons and evaluations of current stateof-the-art SSL methods (namely MOCO, BYOL, SCRL) are conducted and reported for the object detection task.
Kanacı, AytaçTeeti, Izzedin Bradley, AndrewCuzzolin, Fabio
School of Engineering, Computing and Mathematics
Year of publication: 2022Date of RADAR deposit: 2022-10-19