Unmasking Anomalies in Road-Scene Segmentation

Abstract

Anomaly segmentation is a critical task for driving applications, and it is approached traditionally as a per-pixel classification problem. However, reasoning individually about each pixel without considering their contextual semantics results in high uncertainty around the objects' boundaries and numerous false positives.
We propose a paradigm change by shifting from a per-pixel classification to a mask classification. Our mask-based method, Mask2Anomaly, demonstrates the feasibility of integrating an anomaly detection method in a mask-classification architecture. Mask2Anomaly includes several technical novelties that are designed to improve the detection of anomalies in masks: i) a global masked attention module to focus individually on the foreground and background regions; ii) a mask contrastive learning that maximizes the margin between an anomaly and known classes; and iii) a mask refinement solution to reduce false positives. Mask2Anomaly achieves new state-of-the-art results across a range of benchmarks, both in the per-pixel and component-level evaluations. In particular, Mask2Anomaly reduces the average false positives rate by 60% wrt the previous state-of-the-art.

About the speaker

Shyam Nandan Rai is a doctoral student in the VANDAL lab at Politecnico di Torino under the supervision of Prof. Barbara Caputo and co-supervision of Prof. Zeynep Akata. His current areas of interest lie in the amalgamation of semantic segmentation, open-world detection, and federated learning.
He completed his master at IIIT Hyderabad, where he was under the joint supervision of Prof. C.V. Jawahar, Prof. Vineeth N Balasubramanian, and Dr. Anbumani Subramanian on the problems of image restoration in adverse weather conditions.