Modeling Camera Effects to Improve Visual Learning from Synthetic Data
This addresses the challenge of limited generalizability in visual learning from synthetic data for autonomous driving applications, but it is incremental as it builds on existing work by focusing on sensor variations.
The paper tackles the problem of domain gap between synthetic and real images for object detection in urban driving scenes by proposing an automatic augmentation pipeline to model sensor effects like chromatic aberration and noise, resulting in improved performance when tested on real data.
Recent work has focused on generating synthetic imagery to increase the size and variability of training data for learning visual tasks in urban scenes. This includes increasing the occurrence of occlusions or varying environmental and weather effects. However, few have addressed modeling variation in the sensor domain. Sensor effects can degrade real images, limiting generalizability of network performance on visual tasks trained on synthetic data and tested in real environments. This paper proposes an efficient, automatic, physically-based augmentation pipeline to vary sensor effects --chromatic aberration, blur, exposure, noise, and color cast-- for synthetic imagery. In particular, this paper illustrates that augmenting synthetic training datasets with the proposed pipeline reduces the domain gap between synthetic and real domains for the task of object detection in urban driving scenes.