IPAB Workshop - 31/03/2022

Speaker: Panagiotis Eustratiadis

Title: Attacking Adversarial Defences by Smoothing the Loss Landscape

Abstract: We investigate a family of methods for defending against adversarial attacks that owe part of their success to creating a rugged loss landscape that adversaries find difficult to navigate. A common, but not universal, way to achieve this effect is via the use of stochastic neural networks. We show that this is a form of gradient obfuscation, and propose a general extension to gradient-based adversaries based on the Weierstrass transform, which smooths the surface of the loss function and provides more reliable gradient estimates. We further show that the same principle can strengthen gradient-free adversaries. We demonstrate the efficacy of our loss-smoothing method against both stochastic and non-stochastic adversarial defences that exhibit robustness due to this type of obfuscation. Furthermore, we provide analysis of how it interacts with Expectation over Transformation; a popular gradient-sampling method currently used to attack stochastic defences.

Speaker:  Linxu Fan

Title: Simulating brittle fracture with material points

Abstract:  Large-scale topological changes play a key role in capturing the fine de-bris of fracturing virtual brittle material. Real-world, tough brittle fractures have dynamic branching behaviour but numerical simulation of this phe-nomena is notoriously challenging. In order to robustly capture these visual characteristics, we simulate brittle fracture by combining elastodynamic con-tinuum mechanical models with rigid-body methods: A continuum damage mechanics problem is solved, following rigid-body impact, to simulate crack propagation by tracking a damage field. We combine the result of this elas-tostatic continuum model with a novel technique to approximate cracks as a

non-manifold mid-surface, which enables accurate and robust modelling of material fragment volumes to compliment fast-and-rigid shatter effects. For enhanced realism, we add fracture detail, incorporating particle damage-time to inform localised perturbation of the crack surface with artistic control. We evaluate our method with numerous examples and comparisons, showing that it produces a breadth of brittle material fracture effects and with low sim-ulation resolution to require much less time compared to fully elastodynamic simulations.

Speaker: Arushi Goel

Title: Scene Graph Generation - A structured and holistic representation of Images

Abstract:  Scene graph generation (SGG) aims to capture a wide variety of interactions between pairs of objects, which is essential for full scene understanding.  Existing SGG methods fail to acquire complex reasoning about visual and textual correlations due to various biases in training data. In this talk, I will discuss a novel framework for SGG training that exploits relation labels based on their informativeness. 

Our model-agnostic training procedure imputes missing informative relations for less informative samples in the training data and trains a SGG model on the imputed labels along with existing annotations. This approach can successfully be used in conjunction with state-of-the-art SGG methods and improves their performance significantly in multiple metrics on the popular Visual Genome benchmark for studying scene graph generation. Furthermore, we also obtain considerable improvements for unseen triplets in a more challenging zero-shot setting.