Semantic Segmentation

Single-Task Multi-Decoder Network Ensembles

Investigation and implementation of ensemble-based semantic segmentation methods while looking at performance, floating point operations, execution time, and out-of-distribution robustness.

Universal Adversarial Attacks on Videos

Investigation of methods for creating realistically looking universal adversarial attacks for semantic segmentation on video data.

Adversarial Robustness of Teacher-Student Networks

Investigation of methods for teacher-student networks performing semantic segmentation under the context of adversarial robustness.