Polarimetric image augmentation
This work addresses the challenge of autonomous navigation in urban environments by improving segmentation of specular obstacles, though it is incremental as it adapts existing augmentation techniques to a specific data type.
The paper tackled the problem of segmenting specular reflections in urban robotics by proposing a regularized augmentation method for polarimetric images, which are not straightforward to augment like RGB images, resulting in an 18.1% improvement in IoU on real-world data.
Robotics applications in urban environments are subject to obstacles that exhibit specular reflections hampering autonomous navigation. On the other hand, these reflections are highly polarized and this extra information can successfully be used to segment the specular areas. In nature, polarized light is obtained by reflection or scattering. Deep Convolutional Neural Networks (DCNNs) have shown excellent segmentation results, but require a significant amount of data to achieve best performances. The lack of data is usually overcomed by using augmentation methods. However, unlike RGB images, polarization images are not only scalar (intensity) images and standard augmentation techniques cannot be applied straightforwardly. We propose to enhance deep learning models through a regularized augmentation procedure applied to polarimetric data in order to characterize scenes more effectively under challenging conditions. We subsequently observe an average of 18.1% improvement in IoU between non augmented and regularized training procedures on real world data.