Latest events

AIML Research Seminar: Creating a Video Narrative From a Text-based Screen Play

Miguel Balbin — Tue, 01 Jul 2025 14:43:21 +0930

In this session, Sam Hodge, AIML Machine Learning Engineer, explains the process of using synthetic actors to perform dialog specified in a screen play using a combination of audio and 3D speech synthesis and then overcoming the “uncanny valley” sensation through stylisation with ongoing work toward photorealism.

AIML Research Seminar: Protecting SAR Imagery from Adversarial Attack Using a Diffusion Model

Miguel Balbin — Wed, 16 Jul 2025 11:06:25 +0930

Efficient wide-area surveillance of the maritime environment requires automated detection and classification techniques. Deep learning can be exploited for this purpose, but it is susceptible to adversarial attacks that can cause misclassifications. Diffusion models have previously been shown to improve robustness against adversarial perturbations in electro-optic imagery. This work investigates how such adversarial defence techniques perform against white-box attacks on a high-resolution Synthetic Aperture Radar (SAR) dataset. Using a maritime SAR dataset, we demonstrate the benefits of applying a diffusion model, including improved classification performance and reduced sea clutter.

AIML Connect Fridays: Coded Bias Screening

Miguel Balbin — Tue, 22 Jul 2025 09:52:37 +0930

AIML and the Lot Fourteen Innovation District presented a screening of Coded Bias, a documentary about artificial intelligence and the biases that can be embedded within. When MIT Media Lab researcher Dr Joy Buolamwini discovers that facial recognition does not see dark-skinned faces accurately, she worked to find out why these systems failed. After discovering that facial recognition programs only worked for her when she wore an actual white mask, she went on to research how else tech systems can affect minorities and embarks on a journey to push for the first-ever U.S. legislation against bias in algorithms that impact us all.