Chapter
The Challenges of Uncurated Data for Self-Supervised Learning
1:15:27 - 1:21:17 (05:49)
The use of uncurated data for self-supervised learning presents challenges due to the inherent biases of photographers, and the reliance on data augmentation techniques designed for ImageNet.
Clips
Researchers have discovered that very large models can be trained in a self-supervised manner on unfiltered internet images, which is made possible by the use of a data augmentation method designed for self-supervised learning in vision that is overfitted to ImageNet.
1:15:27 - 1:19:22 (03:54)
Self-supervised Learning
Summary
Researchers have discovered that very large models can be trained in a self-supervised manner on unfiltered internet images, which is made possible by the use of a data augmentation method designed for self-supervised learning in vision that is overfitted to ImageNet.
ChapterThe Challenges of Uncurated Data for Self-Supervised Learning
Episode#206 – Ishan Misra: Self-Supervised Deep Learning in Computer Vision
PodcastLex Fridman Podcast
The idea of uncurated data is not fully accurate as there are biases present in uncurated data collection, such as photographer's bias, and not all parts of the world have equal access to the internet which would skew the data.
1:19:22 - 1:21:17 (01:55)
Uncurated data
Summary
The idea of uncurated data is not fully accurate as there are biases present in uncurated data collection, such as photographer's bias, and not all parts of the world have equal access to the internet which would skew the data.