Physically-admissible polarimetric data augmentation for road-scene analysis
This work addresses the robustness issue in polarimetric imaging for road-scene analysis, offering a domain-specific solution that is incremental by adapting existing generative models to enforce physical constraints.
The paper tackles the problem of limited training data for polarimetric imaging in road-scene analysis by proposing a physically-admissible data augmentation method using CycleGAN, which improves object detection performance by up to 9% for cars and pedestrians.
Polarimetric imaging, along with deep learning, has shown improved performances on different tasks including scene analysis. However, its robustness may be questioned because of the small size of the training datasets. Though the issue could be solved by data augmentation, polarization modalities are subject to physical feasibility constraints unaddressed by classical data augmentation techniques. To address this issue, we propose to use CycleGAN, an image translation technique based on deep generative models that solely relies on unpaired data, to transfer large labeled road scene datasets to the polarimetric domain. We design several auxiliary loss terms that, alongside the CycleGAN losses, deal with the physical constraints of polarimetric images. The efficiency of this solution is demonstrated on road scene object detection tasks where generated realistic polarimetric images allow to improve performances on cars and pedestrian detection up to 9%. The resulting constrained CycleGAN is publicly released, allowing anyone to generate their own polarimetric images.