Explore cutting-edge research in visual scene understanding and grounded language comprehension in this 1-hour 22-minute talk by Kevin Murphy from Google Research. Delve into topics such as semantic segmentation, object detection, instance segmentation, and person detection/pose estimation, including award-winning systems like DeepLab and entries in the COCO'16 competition. Discover work on visually grounded referring expressions, discriminative image captioning, and generative models of visual imagination. Learn how these components can be integrated to create systems that better comprehend images and words, advancing the field of AI and machine learning. Gain insights from Murphy's extensive experience in computer science, statistics, and machine learning, spanning academia and industry.