I am thrilled to announce the preprint “Uncertain but useful: Leveraging cnn variability into data augmentation“ is available on arXiv.

Description

This work explores how variability in convolutional neural network inference can be turned into a useful signal for data augmentation, with the goal of improving model robustness and generalization.

Authors

  • Ines Gonzalez-Pepe
  • Vinuyan Sivakolunthu
  • Yohan Chatelain
  • Tristan Glatard