Self-Supervised Domain Mismatch Estimation for Autonomous Perception
This addresses the need for self-awareness in autonomous perception systems, though it is incremental as it builds on existing autoencoder and domain adaptation techniques.
The paper tackles the problem of monitoring semantic segmentation performance in autonomous driving by proposing a self-supervised autoencoder that estimates domain mismatch using a novel metric based on earth mover's distance between PSNR distributions, showing strong rank order correlation with segmentation performance.
Autonomous driving requires self awareness of its perception functions. Technically spoken, this can be realized by observers, which monitor the performance indicators of various perception modules. In this work we choose, exemplarily, a semantic segmentation to be monitored, and propose an autoencoder, trained in a self-supervised fashion on the very same training data as the semantic segmentation to be monitored. While the autoencoder's image reconstruction performance (PSNR) during online inference shows already a good predictive power w.r.t. semantic segmentation performance, we propose a novel domain mismatch metric DM as the earth mover's distance between a pre-stored PSNR distribution on training (source) data, and an online-acquired PSNR distribution on any inference (target) data. We are able to show by experiments that the DM metric has a strong rank order correlation with the semantic segmentation within its functional scope. We also propose a training domain-dependent threshold for the DM metric to define this functional scope.