
A novel Multi-Resolution Feature Perturbation (MRFP) technique is proposed to enhance the generalizability of deep semantic segmentation models trained on synthetic data to unseen real-world domains.


coremsg

learning-generalizable-semantic-segmentation-from-simulation-to-real-world-domains-with-multi-resolution-feature-perturbation


Learning Generalizable Semantic Segmentation from Simulation to Real-World Domains with Multi-Resolution Feature Perturbation
