Scene Text Recognition with Permuted Autoregressive Sequence Models (ECCV 2022)
Pretrained weights for the best-performing models. Training datasets: COCO, RCTW17, Uber, ArT, LSVT, MLT19, ReCTS, TextOCR, and OpenVINO. See Appendix F for details.
NOTES:
parseq-tiny
were trained using a slightly different training setup.parseq_small_patch16_224
is not part of the official release, but added here for those who want to finetune on 224x224 px images.