Addressing Model Vulnerability to Distributional Shifts over Image Transformation Sets
This addresses the problem of model robustness for computer vision applications, but it appears incremental as it builds on existing ideas of data augmentation and optimization.
The paper tackles the vulnerability of computer vision models to distributional shifts by formulating a combinatorial optimization problem to identify vulnerable image regions and embedding this into a training procedure with targeted data augmentation. The result is models that are more robust against content-preserving manipulations and distributional shifts, as shown in empirical evaluations on classification and semantic segmentation tasks.
We are concerned with the vulnerability of computer vision models to distributional shifts. We formulate a combinatorial optimization problem that allows evaluating the regions in the image space where a given model is more vulnerable, in terms of image transformations applied to the input, and face it with standard search algorithms. We further embed this idea in a training procedure, where we define new data augmentation rules according to the image transformations that the current model is most vulnerable to, over iterations. An empirical evaluation on classification and semantic segmentation problems suggests that the devised algorithm allows to train models that are more robust against content-preserving image manipulations and, in general, against distributional shifts.