Improving Data and Modeling for Computer Vision Applications

Explore the latest advancements in computer vision for data-centric image classification model development in this 28-minute conference talk by ML Research Scientist Ravi Teja Mullapudi from Snorkel AI. Discover how visual prompts and fast parameter-efficient models built on foundation models provide immediate feedback for rapid iteration on data quality and model performance, leading to significant time-savings and improvements. Learn about adapting model representations through large-scale fine-tuning on weakly labeled data to address limitations of fast but small models trained on fixed features. Gain insights into scaling and model adaptations necessary for transitioning from image-level classification to object-level detection and segmentation. Understand how computer vision data and models can be effectively improved in tandem and adjusted for downstream applications, providing valuable knowledge for enhancing data and modeling techniques in computer vision applications.