About this document

Enhanced Robustness in Image Classification Through DistortionMix A Hybrid Distortion-Based Augmentation Technique by abel bela is a document available to read on EtoBox.

This paper presents DistortionMix, a hybrid distortion-based augmentation technique aimed at enhancing the robustness of image classification models against common corruptions. By applying various distortions such as contrast variation and Gaussian noise during training, DistortionMix significantly improves model performance on corrupted datasets like CIFAR-10-C, achieving up to a 13.8% increase in accuracy on corrupted data while maintaining clean accuracy. The study evaluates multiple architectures, with

Author

abel bela

Language

EN