A Simple Way to Make Neural Networks Robust Against Diverse Image Corruptions

By Evgenia Rusak, Lukas Schott, Roland S. Zimmermann, Julian Bitterwolf, Oliver Bringmann, Matthias Bethge, Wieland Brendel. 2020

243 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Zhe Gan
2. Yoshua Bengio
3. Trevor Darrell
4. Mohit Bansal
5. Jianfeng Gao
6. Jingjing Liu
7. Linjie Li
8. Sergey Levine

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Trevor Darrell
3. Dacheng Tao
4. Ming-Hsuan Yang
5. Xiaogang Wang
6. Cordelia Schmid
7. Jitendra Malik
8. Andrew Zisserman

In Safety, Robustness & Interpretability

1. Zhe Gan
2. Shuaicheng Niu
3. George Pappas
4. Peilin Zhao
5. Alejandro Ribeiro
6. Trevor Darrell
7. Hugo Larochelle
8. Yaofo Chen

Citations in the corpus

• An Effective Anti-Aliasing Approach for Residual Networks (2020, methodology): “1% for [15].”
• MedViT: A Robust Vision Transformer for Generalized Medical Image Classification (2023, methodology): “These include carefully designed model [35, 36], strong data augmentation [37, 38], searched network architecture [39, 40], improved training strategy [41–43], pruning [44] of the weights, and quantization [45], activation functions [46] or better pooling [47, 48].”
• Adversarial Robustness with Semi-Infinite Constrained Learning (2021, background): “To quote from [85]: Data augmentation with Gaussian ... noise serves as a simple yet very strong baseline that is sufficient to surpass almost all previously proposed defenses against common corruptions.”
• Tent: Fully Test-Time Adaptation by Entropy Minimization (2021, methodology): “Tent improves on the prior state-of-the-art, adversarial noise training (Rusak et al., 2020), by fully test-time adaptation without altering training.”
• Efficient Test-Time Model Adaptation without Forgetting (2022, methodology): “…to enable it to cover the possible shift that might be encountered at test time, such as adversarial training strategies (Wong et al., 2020; Rusak et al., 2020; Madaan et al., 2021), various data augmentation techniques (Lim et al., 2019; Hendrycks et al., 2020; Li et al., 2021; Hendrycks…”
• A Closer Look at the Robustness of Vision-and-Language Pre-trained Models (2020, methodology): “Specifically, MUTANT [17] for VQA-CP v2 is trained with excessive additional image-question pairs designed to promote positive bias; while NSM [26] for GQA takes advantage of additional scene graph annotations, which are only provided in GQA.”