Seminars | Adelaide CSIT Graduate Research

All upcoming

Special session
27 Aug 2026 · 14:30–15:30

Speaker:

Dr Na Zhao, Singapore University of Technology and Design

Date and Time:

Thursday 27 August 2026, 14:30–15:30 (Adelaide)

Title:

Beyond the 3D Data Bottleneck: Foundation Models as Teachers, Annotators, and Reasoners

Location:

CE-AIML-G.42 Atrium

Abstract:

3D scene understanding and reasoning are limited less by architectures than by supervision. 3D annotation is slow and expensive, and the datasets we have remain orders of magnitude smaller than the image–text corpora that gave 2D vision and language models their open-world competence. The productive question, then, is what to borrow from foundation models, and how to route it into 3D. This talk traces that question through a series of our recent works, organized around a single observation in embodied AI. One ordinary instruction, such as "open the bottom drawer of the wooden cabinet with the flower vase on top", inherently demands four distinct capabilities: 1) recognizing categories outside the training vocabulary; 2) resolving which instance a description refers to; 3) reasoning about spatial arrangement; and 4) inferring how a part of an object can be acted upon. Across these works, 2D foundation models take on three roles. As teachers, they supply image-wise guidance and vision-language alignment for open-vocabulary 3D detection. As annotators, they synthesize diverse text–3D pairs that any existing 3D grounding method can consume. As reasoners, they operate directly over rendered viewpoints, replacing 3D-specific training with task-driven view selection to enable both spatial reasoning and fine-grained embodied reasoning, including predicting the location of affordance elements and how they move. Together, these works show which role a foundation model can usefully play in a given 3D task depending on where the bottleneck lies. I will close by looking ahead to instruction-driven embodied agents, and to the open challenges that stand in the way.

Bio:

Dr. Na Zhao is an Assistant Professor at the Singapore University of Technology and Design (SUTD), where she directs the Intelligent Machine Perception Lab (IMPL). She received her PhD in Computer Science from the National University of Singapore, where her thesis was awarded the 2021 IMDA Excellence Prize for Best PhD Thesis. Her research spans 3D computer vision, machine learning, and embodied AI, with the goal of building trustworthy autonomous systems that can perceive, reason, and act reliably in real-world applications, including robotics and autonomous driving. She has published 70+ papers in leading venues, including CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, AAAI, ICRA, IROS, IJCV, and TIP, and has attracted competitive research funding totaling over S$14M, including approximately S$4M as lead PI. She serves as an Associate Editor of IEEE Transactions on Circuits and Systems for Video Technology and Knowledge-Based Systems, and as an Area Chair or Senior Program Committee member for venues including NeurIPS, ICLR, ACM MM, AAAI, and IJCAI. She is General Co-Chair of the 33rd International Conference on Multimedia Modeling (MMM 2027), Diversity, Equity, and Inclusion Chair of the 36th ACM Web Conference (WWW 2027), and will serve as Technical Program Co-Chair of the 18th ACM International Conference on Multimedia Retrieval (ICMR 2028).

Online:

Join online
Special session
8 Sep 2026 · 14:30–15:30

Speaker:

Dr Dino Sejdinovic (introduction); Erdun Gao (presentation)

Date and Time:

Tuesday 8 September 2026, 14:30–15:30 (Adelaide)

Title:

AI Safety & Robustness / Trusted Autonomous Systems

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk titles TBC.
Special session
22 Sep 2026 · 14:30–15:30

Speaker:

Dr Arpit Garg

Date and Time:

Tuesday 22 September 2026, 14:30–15:30 (Adelaide)

Title:

AI Safety & Robustness / Trusted Autonomous Systems

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
6 Oct 2026 · 14:30–15:30

Speaker:

Prof Lyle Palmer

Date and Time:

Tuesday 6 October 2026, 14:30–15:30 (Adelaide)

Title:

Probabilistic & Statistical Machine Learning

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
20 Oct 2026 · 14:30–15:30

Speaker:

Dr Henry Li

Date and Time:

Tuesday 20 October 2026, 14:30–15:30 (Adelaide)

Title:

Probabilistic & Statistical Machine Learning

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
3 Nov 2026 · 14:30–15:30

Speaker:

Professor Tat-Jun Chin

Date and Time:

Tuesday 3 November 2026, 14:30–15:30 (Adelaide)

Title:

AI for Space

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
17 Nov 2026 · 14:30–15:30

Speaker:

Alex Mundy

Date and Time:

Tuesday 17 November 2026, 14:30–15:30 (Adelaide)

Title:

Physics Informed Neural Networks

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
1 Dec 2026 · 14:30–15:30

Speaker:

Prof Simon Lucey

Date and Time:

Tuesday 1 December 2026, 14:30–15:30 (Adelaide)

Title:

AIML 2026 Year in Review

Location:

AIML Atrium

Abstract:

AIML Research Seminar — 2026 Year in Review with Prof Simon Lucey, Director.

Candidature review

No seminars in this section.

Thesis defence (public seminar)

No seminars in this section.

GRS seminar

View full 2026 GRS series program (table) →

No seminars in this section.

Special session

Special session
27 Aug 2026 · 14:30–15:30

Speaker:

Dr Na Zhao, Singapore University of Technology and Design

Date and Time:

Thursday 27 August 2026, 14:30–15:30 (Adelaide)

Title:

Beyond the 3D Data Bottleneck: Foundation Models as Teachers, Annotators, and Reasoners

Location:

CE-AIML-G.42 Atrium

Abstract:

3D scene understanding and reasoning are limited less by architectures than by supervision. 3D annotation is slow and expensive, and the datasets we have remain orders of magnitude smaller than the image–text corpora that gave 2D vision and language models their open-world competence. The productive question, then, is what to borrow from foundation models, and how to route it into 3D. This talk traces that question through a series of our recent works, organized around a single observation in embodied AI. One ordinary instruction, such as "open the bottom drawer of the wooden cabinet with the flower vase on top", inherently demands four distinct capabilities: 1) recognizing categories outside the training vocabulary; 2) resolving which instance a description refers to; 3) reasoning about spatial arrangement; and 4) inferring how a part of an object can be acted upon. Across these works, 2D foundation models take on three roles. As teachers, they supply image-wise guidance and vision-language alignment for open-vocabulary 3D detection. As annotators, they synthesize diverse text–3D pairs that any existing 3D grounding method can consume. As reasoners, they operate directly over rendered viewpoints, replacing 3D-specific training with task-driven view selection to enable both spatial reasoning and fine-grained embodied reasoning, including predicting the location of affordance elements and how they move. Together, these works show which role a foundation model can usefully play in a given 3D task depending on where the bottleneck lies. I will close by looking ahead to instruction-driven embodied agents, and to the open challenges that stand in the way.

Bio:

Dr. Na Zhao is an Assistant Professor at the Singapore University of Technology and Design (SUTD), where she directs the Intelligent Machine Perception Lab (IMPL). She received her PhD in Computer Science from the National University of Singapore, where her thesis was awarded the 2021 IMDA Excellence Prize for Best PhD Thesis. Her research spans 3D computer vision, machine learning, and embodied AI, with the goal of building trustworthy autonomous systems that can perceive, reason, and act reliably in real-world applications, including robotics and autonomous driving. She has published 70+ papers in leading venues, including CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, AAAI, ICRA, IROS, IJCV, and TIP, and has attracted competitive research funding totaling over S$14M, including approximately S$4M as lead PI. She serves as an Associate Editor of IEEE Transactions on Circuits and Systems for Video Technology and Knowledge-Based Systems, and as an Area Chair or Senior Program Committee member for venues including NeurIPS, ICLR, ACM MM, AAAI, and IJCAI. She is General Co-Chair of the 33rd International Conference on Multimedia Modeling (MMM 2027), Diversity, Equity, and Inclusion Chair of the 36th ACM Web Conference (WWW 2027), and will serve as Technical Program Co-Chair of the 18th ACM International Conference on Multimedia Retrieval (ICMR 2028).

Online:

Join online
Special session
8 Sep 2026 · 14:30–15:30

Speaker:

Dr Dino Sejdinovic (introduction); Erdun Gao (presentation)

Date and Time:

Tuesday 8 September 2026, 14:30–15:30 (Adelaide)

Title:

AI Safety & Robustness / Trusted Autonomous Systems

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk titles TBC.
Special session
22 Sep 2026 · 14:30–15:30

Speaker:

Dr Arpit Garg

Date and Time:

Tuesday 22 September 2026, 14:30–15:30 (Adelaide)

Title:

AI Safety & Robustness / Trusted Autonomous Systems

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
6 Oct 2026 · 14:30–15:30

Speaker:

Prof Lyle Palmer

Date and Time:

Tuesday 6 October 2026, 14:30–15:30 (Adelaide)

Title:

Probabilistic & Statistical Machine Learning

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
20 Oct 2026 · 14:30–15:30

Speaker:

Dr Henry Li

Date and Time:

Tuesday 20 October 2026, 14:30–15:30 (Adelaide)

Title:

Probabilistic & Statistical Machine Learning

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
3 Nov 2026 · 14:30–15:30

Speaker:

Professor Tat-Jun Chin

Date and Time:

Tuesday 3 November 2026, 14:30–15:30 (Adelaide)

Title:

AI for Space

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
17 Nov 2026 · 14:30–15:30

Speaker:

Alex Mundy

Date and Time:

Tuesday 17 November 2026, 14:30–15:30 (Adelaide)

Title:

Physics Informed Neural Networks

Location:

AIML Atrium

Abstract:

AIML Research Seminar. Format: AIML News, a short introduction by a senior academic, a student/postdoc presentation, Q&A, and afternoon tea. Talk title TBC.
Special session
1 Dec 2026 · 14:30–15:30

Speaker:

Prof Simon Lucey

Date and Time:

Tuesday 1 December 2026, 14:30–15:30 (Adelaide)

Title:

AIML 2026 Year in Review

Location:

AIML Atrium

Abstract:

AIML Research Seminar — 2026 Year in Review with Prof Simon Lucey, Director.