Google, IBM & Meta Certificates — All 10,000+ Courses at 40% Off
One annual plan covers every course and certificate on Coursera. 40% off for a limited time.
Get Full Access
Explore a comprehensive explanation of the Big Transfer (BiT) paper, which introduces a pre-trained ResNet model for general visual representation learning. Delve into the intricacies of pre-training large models and learn best practices for fine-tuning on downstream tasks. Discover how BiT achieves strong performance across a wide range of datasets and data regimes, from limited examples to millions of samples. Examine the model's impressive accuracy on various benchmarks, including ILSVRC-2012, CIFAR-10, and the Visual Task Adaptation Benchmark. Gain insights into the key components that contribute to high transfer performance through detailed analysis. Follow along as the video covers introduction, models, ablation studies, group normalization, weight standardization, visual task adaptation benchmarks, and potential mistakes in implementation.
