Leveraging Active Perception for Improving Embedding-based Deep Face Recognition

Abstract

Even though recent advances in deep learning (DL) led to tremendous improvements for various computer and robotic vision tasks, existing DL approaches suffer from a significant limitation: they typically ignore that robots and cyberphysical systems are capable of interacting with the environment in order to better sense their surroundings. In this work we argue that perceiving the world through physical interaction, i.e., employing active perception, allows for both increasing the accuracy of DL models, as well as for deploying smaller and faster models. To this end, we propose an active perception-based face recognition approach, which is capable of simultaneously extracting discriminative embeddings, as well as predicting in which direction the robot must move in order to get a more discriminative view. To the best of our knowledge, we provide the first embedding-based active perception method for deep face recognition. As we experimentally demonstrate, the proposed method can indeed lead to significant improvements, increasing the face recognition accuracy up to 9%, as well as allowing for using overall smaller and faster models, reducing the number of parameters by over one order of magnitude.

Pseudo-Active Vision For Improving Deep Visual Perception Through Neural Sensory Refinement

“…It is worth noting that a camera can be controlled both regarding its external parameters, e.g., pan and tilt, as well as some of its internal parameters, e.g., exposure and color profile, etc. Even through there is an increasing amount of literature for handling the former [4,5,6,7,8], less focus has been given to the latter (with respect to the performance of DL models). Indeed, most DL algorithms implicitly assume that the heavy pre-processing that is involved most digital camera sensors, e.g., color constancy algorithms [9,10,11,12], will mitigate the effect of varying illumination conditions.…”