We introduce AugLy, a data augmentation library with a focus on adversarial robustness. AugLy provides a wide array of augmentations for multiple modalities (audio, image, text, & video). These augmentations were inspired by those that real users perform on social media platforms, some of which were not already supported by existing data augmentation libraries. AugLy can be used for any purpose where data augmentations are useful, but it is particularly well-suited for evaluating robustness and systematically generating adversarial attacks. In this paper we present how AugLy works, benchmark it compared against existing libraries, and use it to evaluate the robustness of various state-of-the-art models to showcase AugLy's utility. The AugLy repository can be found at https://github.com/facebookresearch/AugLy.

Data augmentation is an important part when the dataset we are using does not contain much information so we cannot use this data alone to make a model out of it because the model will not be generalized due to lack of information in the training data. Let's try to understand this by an example. Let say that we are trying to build an image classification model. The dataset we are using contains 10 classes and 100 images for each class. Now we can build the model, but the question is will it be generalized and optimized enough to use it for the prediction of a new dataset.