Latest events

AIML Connect Fridays: Towards the 3D Human Foundation Agent

Date: Fri, 13 Dec 2024, 3:30 pm - 4:30 pm
Location: AIML

This talk will describe current progress on building a 3D Human Foundation Agent (HFA) that can perceive the world and the humans in it. The HFA is a digitally embodied agent that understands human behavior and responds to it using its “motor system” to translate its goals into 3D actions. The Human Foundation Agent must (1) perceive human movement in 3D, (2) understand the goals, implications, and emotions inherent in that movement, and (3) plan and generate natural motor activity to (4) drive a digital or physical embodiment that interacts with real or virtual humans in real or virtual 3D worlds. This talk will focus on current progress and the path to building HFAs through 3D human motion capture from video, synthetic training data, generative behaviour modelling, AI-driven graphics, and large vision-language models that are fine-tuned to understand 3D humans. HFAs will radically change how people interact with machines. So much so that a child born today will have trouble imagining a world in which technology doesn’t understand their motions and behaviours.

AIML Research Seminar: What Makes AI Safe for Human Decision-Making?

Date: Tue, 10 Dec 2024, 10:30 am - 11:15 am
Location: AIML

Abstract: Lana is a cognitive psychology student at AIML investigating medical decision-making with AI. She incorporates knowledge of human factors, expertise, and cognitive science with areas of bioethics and AI safety. She will cover the work of her PhD, findings, and future directions.

AIML Special Presentation: From Instructional Diagrams to Real-World Assembly

Date: Thu, 5 Dec 2024, 10:30 am - 11:15 am
Location: AIML

Abstract: From instructional diagrams to real-world assembly, understanding and bridging the gap between diagrammatic instructions and practical actions is a significant challenge. To address this, we introduce the Ikea Assembly in the Wild (IAW) Dataset, which comprises 183 hours of diverse furniture assembly videos and nearly 8,300 corresponding illustrations from assembly manuals, annotated with their ground truth alignments. Leveraging this dataset, we tackle three key tasks: First, we align segments of in-the-wild assembly videos with corresponding instructional diagrams using a supervised contrastive learning approach that captures subtle diagrammatic details. Second, we predict the precise start and end times of the steps outlined in the manuals within the video sequences, enabling accurate temporal grounding. Finally, we demonstrate a furniture assembly pipeline where furniture parts are selected and their 6D poses are predicted based on cues from the instructional diagrams. Our methods significantly advance multimodal alignment and learning for practical assembly, achieving state-of-the-art performance on tasks such as retrieval, temporal grounding, and 3D part assembly.

AIML Special Presentation: Several Recent Results in 4D Reconstruction and Generative Modelling

Date: Mon, 2 Dec 2024, 10:30 am - 11:30 am
Location: AIML

Dr Golyanik’s talk focuses on several recent approaches for generative modelling and 3D reconstruction of non-rigid scenes from various conditioning signals and inputs, such as textual prompts, RGB images, or event streams. He also outlines promising directions for how the field could benefit from modern and upcoming quantum hardware.

AI on the Ground Seminar: Building responsible AI for healthcare: challenges and ensuring fairness

Date: Fri, 29 Nov 2024, 10:30 am - 11:30 am
Location: AIML

Abstract: The presentation covers the use of machine learning in healthcare, focusing on project deployment and key lessons learned. An outline of the Machine Learning Medical Directive project and the underlying research is given first, then the primary challenges that arose during deployment. Solutions to these challenges will be discussed, with special attention to fairness and ethical considerations in healthcare AI.

AIML Research Seminar: A Couple of AI4Space Problems

Date: Tue, 26 Nov 2024, 10:30 am - 11:15 am
Location: AIML

Abstract: We’re living in an exciting new space era, where the cost of ‘going into space’ has dropped significantly, and the reasons for doing so have expanded dramatically. It’s not just about the numbers anymore—we’re tackling highly sophisticated tasks that once existed only in sci-fi novels. This surge in scale and complexity is why we turn to AI.

AIML Visits Roseworthy Campus

Date: Fri, 22 Nov 2024, 10:00 am - 4:30 pm
Location: Roseworthy Campus, ��˴�Ƭ

AIML academics and professional staff visited the ��˴�Ƭ's Roseworthy Campus for an engaging day at the School of Animal & Veterinary Sciences. The event aimed to introduce AIML staff to the campus and foster meaningful discussions about the applications of machine learning in veterinary and animal sciences.

AIML Special Presentation: Convergence and Asymptotic Optimality of the Heavy Ball Method and Its Relatives

Date: Wed, 20 Nov 2024, 2:30 pm - 3:30 pm
Location: AIML

Abstract: In this talk we first aim to shed light on the urban legend surrounding the ‘complexity lower bound’ for the heavy ball algorithm. Second, we revisit the original heavy-ball algorithm proposed by Polyak and provide a conditions for it to be globally converging, and provide step-size rules. Then, we investigate the performance of a related algorithm dubbed the Accelerated Generalised Gradient Method and see how how it can be beneficial in the case of tracking the minimum of a time-varying function.

AIML Special Presentation: Artificial Intelligence for Autonomous Scientific Exploration

Date: Mon, 18 Nov 2024, 10:30 am - 11:30 am
Location: AIML

David Wettergreen creates robots that explore and conducts field experiments in polar climates, deserts, underwater caverns, and volcanic craters. He has led numerous research projects including a decade of robotic investigation of microbial life in the Atacama Desert. His work in science autonomy enables robotic explorers to detect, classify, and evaluate geologic and biologic features to autonomously interpret and act upon their scientific observations. This work applies to space exploration and to applications in agriculture, forestry, ecology, and marine science.

AIML Special Presentation: From Multi-view 3D CT Volume Reconstruction to 3D Content Generation

Date: Thu, 31 Oct 2024, 10:30 am - 11:30 am
Location: AIML

Abstract: Reconstructing the geometry of 3D geometry from multi-view images or multiple projections has been a fundamental task in Computer Vision and Medical Imaging. The recent emerging generative-AI technology also offers new capability to generate realistic 3D contents simply from user natural input. In this talk, I will report two of our recent works on 3D reconstruction and 3D content generation. The first work is about how to reconstruct 3D CT volumetric shape from X-ray/CT projections, and the second one is on 3D content creation using pre-trained large-scale diffusion models such as the Stable Diffusion. For the latter, I will cover both 3D single object generation, as well as 3D scene generation from hand drawn sketches.

��˴�Ƭ