BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Australia/Canberra BEGIN:DAYLIGHT DTSTART:20261004T030000 TZOFFSETFROM:+1000 TZOFFSETTO:+1100 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=10 TZNAME:AEDT END:DAYLIGHT BEGIN:STANDARD DTSTART:20260405T020000 TZOFFSETFROM:+1100 TZOFFSETTO:+1000 RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=4 TZNAME:AEST END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260412T080210Z UID:4751FC14-C925-47FF-BE4B-EF93A4C89BAD DTSTART;TZID=Australia/Canberra:20260409T210000 DTEND;TZID=Australia/Canberra:20260409T220000 DESCRIPTION:Latent diffusion models underly modern image generation\, which requires a variational auto-encoder (VAE) for image encoding and decoding \, and a diffusion transformer for generation. While end-to-end training h as been the spirit of deep learning\, it is surprising that latent diffusi on models are not trained end-to-end\, causing representation bottlenecks. In this talk\, I will introduce our work that jointly trains the VAE and diffusion transformer and show how it accelerates training and yields high quality images. Further\, I will discuss use cases where the resulting en d-to-end trained VAEs bring significant benefits. This includes higher-qua lity text-to-image generation and automatic agentic search of diffusion tr ansformer architectures. I will conclude with new perspectives.\n\nLiang Z heng\n\nAustralian National University\n\nDr. Liang Zheng is an Associate Professor at the Australian National University and a Research Scientist a t Canva. He is interested in representation learning for perception and ge neration. He contributed many useful datasets and methods to the object re -identification field that were later used in wider domains. He is current ly working on image generation in both aspects of pre-training and post-tr aining. He is a Program Chair for ACM MM’24\, MM’28\, andAVSS'24\, and a General Chair for AVSS’27 and DICTA 2027. He is a regular area chair for important conferences and an Associate Editor for TPAMI. He has bachel or degrees in Biology\, Economics and a PhD degree in Computer Science fro m Tsinghua University.\n\nVirtual: https://events.vtools.ieee.org/m/551957 LOCATION:Virtual: https://events.vtools.ieee.org/m/551957 ORGANIZER:ambarish.natu@gmail.com SEQUENCE:10 SUMMARY:Image generation with end-to-end training and benefits of a good VA E URL;VALUE=URI:https://events.vtools.ieee.org/m/551957 X-ALT-DESC:Description: <br /><p>Latent diffusion models underly modern ima ge generation\, which requires a variational auto-encoder (VAE) for image encoding and decoding\, and a diffusion transformer for generation. While end-to-end training has been the spirit of deep learning\, it is surprisin g that latent diffusion models are not trained end-to-end\, causing repres entation bottlenecks. In this talk\, I will introduce our work that jointl y trains the VAE and diffusion transformer and show how it accelerates tra ining and yields high quality images. Further\, I will discuss use cases w here the resulting end-to-end trained VAEs bring significant benefits. Thi s includes higher-quality text-to-image generation and automatic agentic s earch of diffusion transformer architectures. I will conclude with new per spectives.</p>\n<p>&nbsp\;</p>\n<h4 class="ant-typography css-1daqxe2 css- var-_r_0_">Liang Zheng</h4>\n<h4 class="ant-typography css-1daqxe2 css-var -_r_0_"><strong>Australian National University</strong></h4>\n<p><strong>D r. Liang Zheng is an Associate Professor at the Australian National Univer sity and a Research Scientist at Canva. He is interested in representation learning for perception and generation. He contributed many useful datase ts and methods to the object re-identification field that were later used in wider domains. He is currently working on image generation in both aspe cts of pre-training and post-training. He is a Program Chair for ACM MM&rs quo\;24\, MM&rsquo\;28\, andAVSS'24\, and a General Chair for AVSS&rsquo\; 27 and DICTA 2027. He is a regular area chair for important conferences an d an Associate Editor for TPAMI. He has bachelor degrees in Biology\, Econ omics and a PhD degree in Computer Science from Tsinghua University.</stro ng></p> END:VEVENT END:VCALENDAR