Auke Wiggers, an AI research scientist at Qualcomm, joins us today. We talked with Auke about his team’s recent research on data compression using generative models. We discuss the connection between historical compression research and the current trend of neural compression, as well as the advantages of neural codecs that learn to compress data from examples. We also investigate the performance evaluation process and recent developments demonstrating that these models can operate in real-time on a mobile device. Finally, we discuss another ICLR paper, “Transformer-based transform coding,” which proposes a vision transformer-based architecture for image and video coding, as well as some of his team’s other conference-accepted works.
