Embedded Files
Prof. Aylin Yener (Fellow, IEEE), The Ohio State University, USA
Prof. Aylin Yener (Fellow, IEEE), The Ohio State University, USA
Speech Title: The Role of Pre-trained Models and Generative AI in Semantic Communication.
Abstract: TBD.
Prof. Deniz Gunduz (Fellow, IEEE), Imperial College London, UK
Prof. Deniz Gunduz (Fellow, IEEE), Imperial College London, UK
Speech Title: Generative AI for Wireless Content Delivery
Abstract: Over the past few years, significant advancements have taken place in the field of wireless image/video delivery, thanks to the utilization of deep neural networks (DNNs) for joint source-channel coding. This approach, known as Deep Joint Source-Channel Coding (DeepJSCC), operates in an end-to-end fashion, eliminating the need for separate blocks for compression, channel coding, modulation, and power allocation, and is shown to outperform conventional separation-based approaches. Concurrently, the domain of generative AI has made significant strides, garnering attention for its impressive applications in photorealistic image generation and conversational agents. In this talk, I will showcase how DeepJSCC can be combined with powerful pretrained generative models to enable a notable enhancement in the quality of reconstructed images at the receiver. By leveraging Generative DeepJSCC, which incorporates generative adversarial networks (GANs) or potent diffusion models, it becomes possible to produce realistic images with high fidelity, even under extreme conditions of limited bandwidth and low signal-to-noise ratio. This novel communication paradigm opens up exciting possibilities such as enabling wireless gaming and enriching metaverse experiences.
Page updated
Google Sites
Report abuse