2022 Seminars


  • 07 July 2022 (10.00am, Thursday, One Tree Hill) Dr Wei Qi YAN (CS, AUT) Topics on “CS Retreat: The Past, Present and Future of CeRV”.
  • 8 June 2022 – 29 June 2022 (Wednesday, 5pm, Online) (Wei Qi Yan, CS AUT) Topics on “Deep Learning and  Deep Neural Networks
  • (1) Introduction to deep learning (8 June 2022)
  • (2) PG programming for deep learning (15 June 2022)
  • (3) Manifold learning (22 June 2022)
  • (4) Reinforcement learning (29 June 2022)
  • 26 April 2022 (5.00pm, Tuesday, Online) Mr Kasun Moolika Gedara (CS, AUT) Topics on “Computational Methods for Visual Blockchain in Intelligent Surveillance
  • Visual blockchain takes the novel idea of how to manage and export precise data from surveillance systems for smart cities. Intelligent surveillance automatically provides solid and trustable evidence from networked sensors which are deployed in smart cities. The intelligent systems may suffer limited tamper resistance of the gathered data from digital cameras. The decentralised nature of blockchains has the ability to integrate video frames as the output by using the way of selecting the most appropriate cryptographical algorithms and creating computational methods for visual blockchain so as to proffer sustainable and reliable solutions for the existing problem. In this project, we will devote to deeply investigate computational methods for visual blockchain such as signatures and commitments as well as Merkle tree and secret sharing, our contributions of this project are to: (1) Enrich the security shield of visual blockchain by using computational methods, and (2) create a web-based prototype for implementing visual blockchains so as to minimise the existing gaps. 
  • 31 Mar 2022 – 09 June 2022 (Thursday, 12pm, Online) Topics on “The Latest Progress of Our Deep Learning Projects”
  • (1) Self-supervised depth estimation of traffic scenes using deep learning (X. Liu, 31 Mar 2022)
  • (2) Fruit recognition from digital images using manifold learning (B. Xiao, 07 Apr 2022)
  • (3) Manifold learning for CGN network (F. Younus, 14 Apr 2022)
  • (4) Waste classification from digital images using ConvNet (J. Qi, 21 Apr 2022)
  • (5) Face image inpainting based on GAN net (X. Gao, 28 Apr 2022)
  • (6) Human emotion recognition (R. Alexandre, 05 May 2022)
  • (7) Sailboat detection with improved YOLO Transformer (Z. Luo, 12 May 2022)
  • (8) Sign language recognition from videos using detection Transformer (Y. Liu, 19 May 2022)
  • (9) Human action recognition based on Transformer in deep learning (C. Liang, 26 May 2022)
  • (10) Masked face recognition in real-time using MobleNet (M. Liu, 02 Jun 2022)
  • (11) Traffic sign recognition based on Capsule networks (W. Hao, 09 Jun 2022)
  • (12) Computational methods for visual blockchain in intelligent surveillance (K. Gedara, 16 Jun 2022)
  • (13) ACM ICCCV 2022 Video Presentations (Z. Luo, M. Liu, 23 Jun 2022)
  • (14) ACM ICCCV 2022 Video Presentations (Y. Zhu, C. Liang, 30 Jun 2022)
  • 24 February 2022 (2.00pm, Thursday, Online) Ms Jianchun Qi (CS, AUT) Topics on “A Novel Framework for Waste Classification from Digital Image Using Transformers in Deep Learning”
  • Waste classification is one of most important ways to protect our environment. There are a spate of benefits to waste sorting. For example, it can reduce pollution and protect NZ ecological environment. Besides, it can also cut down land space and improve land utilisation. The recycling of waste can make effective use of resources. It is important that waste is disposed efficiently and cost-effectively, hence, automated waste sorting is one of the solutions. Even though a pretty assortment of deep learning algorithms have been applied to the field of waste classification, there is still much room for improvement in terms of sorting efficiency and detection accuracy. Moreover, a collection of waste data is a current challenge due to a wide variety of waste classes. Therefore, it is necessary to construct an efficient deep learning model and collect sufficient data to improve the efficiency of waste classification. Therefore, we put forward a new method of waste classification by using attention mechanism of deep learning. The purpose is to implement an efficient, accurate, and reliable waste classification method, which is one of the ways to improve environmental protection. Currently, the network we are using is Swin Transformer. 
  • 20 Jan 2022 – 24 Mar 2022 (Thursday, 12pm, Online) (Wei Qi Yan, CS AUT) Topics on “PG Writing, Programming and Mathematics for Deep Learning”
  • (1) PG writing for deep learning (20 January 2022)
  • (2) PG programming for deep learning (27 January 2022)
  • (3) Functional analysis (03 February 2022)
  • (4) Finite fields (10 February 2022)
  • (5) Reinforcement learning (17 February 2022)
  • (6) Mathematical control theory (24 February 2022)
  • (7) Probabilistic graphical models (03 March 2022)
  • (8) Manifold learning (10 March 2022)
  • (9) Graph neural networks (17 March 2022)
  • (10) Transformers (24 March 2022)