Input Dropout for Spatially Aligned Modalities
This addresses a practical limitation in deploying multi-sensor systems for computer vision applications, offering an incremental improvement for scenarios with spatially aligned modalities.
The paper tackles the problem of leveraging additional modalities available in training datasets but not at test time by proposing Input Dropout, a technique that stochastically hides modalities during training to improve performance on tasks like dehazing, object tracking, detection, and classification.
Computer vision datasets containing multiple modalities such as color, depth, and thermal properties are now commonly accessible and useful for solving a wide array of challenging tasks. However, deploying multi-sensor heads is not possible in many scenarios. As such many practical solutions tend to be based on simpler sensors, mostly for cost, simplicity and robustness considerations. In this work, we propose a training methodology to take advantage of these additional modalities available in datasets, even if they are not available at test time. By assuming that the modalities have a strong spatial correlation, we propose Input Dropout, a simple technique that consists in stochastic hiding of one or many input modalities at training time, while using only the canonical (e.g. RGB) modalities at test time. We demonstrate that Input Dropout trivially combines with existing deep convolutional architectures, and improves their performance on a wide range of computer vision tasks such as dehazing, 6-DOF object tracking, pedestrian detection and object classification.