Getting High-Quality Data for Your Computer Vision Models - Building Computer Vision Models

Explore the importance of high-quality data in computer vision models through this 50-minute talk by Iva Gumnishka, founder and CEO of Humans in the Loop. Learn about the shift from model-centric to data-centric AI, focusing on data quality, diversity, and consistency. Discover methods for collecting and annotating datasets, ensuring representativeness, and avoiding harmful biases. Gain insights into algorithmic bias mitigation, self-supervised pretraining, and the future of computer vision data. Understand the EU's perspective on AI and examine canonical large-scale datasets. Perfect for data scientists, AI professionals, and anyone interested in improving computer vision model performance through better data practices.

– Introduction
– Why should we care about high-quality data for computer vision?
– Algorithmic methods for bias mitigation
– Model centric to data centric AI
– Take of EU
– Canonical large-scale dataset
– Self-supervised pretraining
– Image collection and labeling
– Data quality
– The future
– Questions