This repo contains research studies of learning from synthetic data (mostly images), including:
StableRep- StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners. Yonglong Tian*, Lijie Fan*, Phillip Isola, Huiwen Chang, Dilip Krishnan.Scaling- Scaling Laws of Synthetic Images for Model Training ... for Now. Lijie Fan*, Kaifeng Chen, Dilip Krishnan, Dina Katabi, Phillip Isola, Yonglong Tian*.SynCLR- Learning Vision from Models Rivals Learning Vision from Data. Yonglong Tian*, Lijie Fan*, Kaifeng Chen, Dina Katabi, Dilip Krishnan, Phillip Isola.
This is not an officially supported Google product.
Apache2 license.
Yonglong Tian (yonglong@google.com)
Lijie Fan (lijiefan@mit.edu)
