CertViT: Certified Robustness of Pre-Trained Vision Transformers
Nov 09, 2023
Abstract
Lipschitz bounded neural networks are certifiably robust and have a good trade-off between clean and certified accuracy. Existing Lipschitz bounding methods train from scratch and are limited to moderately sized networks (< 6M parameters). They require a fair amount of hyper-parameter tuning and are computationally prohibitive for large networks like Vision Transformers (5M to 660M parameters). Obtaining certified robustness of transformers is not feasible due to the non-scalability and inflexibility of the current methods. This work presents CertViT, a two-step proximal-projection method to achieve certified robustness from pre-trained weights. The proximal step tries to lower the Lipschitz bound and the projection step tries to maintain the clean accuracy of pre-trained weights. We show that CertViT networks have better certified accuracy than state-of-the-art Lipschitz trained networks. We apply CertViT on several variants of pre-trained vision transformers and show adversarial robustness using standard attacks.
ICML Workshop 2023
Contributed by
Kavya Gupta , Sagar Verma
Related Research
Synthetix: Pipeline for Synthetic Geospatial Data Generation
Remote sensing is crucial in various domains, such as agriculture, urban planning, environmental monitoring, and disaster management. However, acquiring real-world remote sensing data can be challenging due to cost, logistical constraints, and privacy concerns. To overcome these limitations, synthetic data has emerged as a promising approach. We present an overview of the use of synthetic data for remote sensing applications.In this regard, we address three conditions that can drastically affect the optimization of computer vision algorithms: lighting conditions, fidelity of the 3D model, and resolution of the synthetic imagery data. We propose a highly configurable pipeline called Synthetix as part of our GeoEngine platform for synthetic data generation. Synthetix allows us to quickly create large amounts of aerial and satellite imagery under varying conditions, given a few samples of 3D objects on real-world scenes. We demonstrate our pipeline’s effectiveness by generating 3D scenes from 35 real-world locations and utilizing these scenes to generate different versions of datasets and answer the three questions. We conduct an in-depth ablation study and show that considering different environments and weather conditions increases the reliability and robustness of the deep learning networks.
Post Wildfire Burnt-up Detection using Siamese UNet
In this article, we present an approach for detecting burnt area due to wild fire in Sentinel-2 images by leveraging the power of Siamese neural networks. By employing a Siamese network, we are able to efficiently encode the feature extraction process for pairs of images. This is achieved by utilizing two branches within the Siamese network, which capture and combine information at different resolutions to make predictions. The weights are shared between these two branches in siamese networks. This design allows to effectively analyze the changes between two remote sensing images, enabling precise identification of areas impacted by forest wildfires in the state of California as part of ChaBuD challenge thereby assisting local authorities in effectively monitoring the impacted regions and facilitating the restoration process. We experimented with various model architectures to train ChaBuD dataset and carefully evaluated the performance. Through rigorous testing and analysis, we have achieved promising results, ultimately obtaining a final private score (IoU) of 0.7495 on the hidden test dataset. The code is available at https://github.com/kavyagupta/chabud. We also deploy the final model as a point solution for anyone to use at https://firemap.io.