search

amazon-science / semimtr-text-recognition

Multimodal Semi-Supervised Learning for Text Recognition (SemiMTR)
Apache License 2.0
81 stars 12 forks source link

Is there an implementation of Seqclr? #2

Closed shantzhou closed 2 years ago

aaberdam commented 2 years ago

Hi, Thanks for the interest in our work!

The major components of SeqCLR are implemented in this codebase. In particular:

The main difference of this codebase with SeqCLR is the base architecture, on which we apply SeqCLR. In the paper, we used SCATTER, and here, we employ ABI-Net. In addition, here, we offer a combined stage of SeqCLR with supervised CE (this can be removed in the config).

shantzhou commented 2 years ago

Hi, Thanks for the interest in our work!

The major components of SeqCLR are implemented in this codebase. In particular:

  • The data loader is implemented in ImageDatasetSelfSupervised.
  • The augmentation pipeline in get_augmentation_pipeline with augmentation_severity=2.
  • The projection module and the instance mapping function are implemented in SeqCLRProj.
  • Lastly, the loss function is implemented in SeqCLRLoss.

The main difference of this codebase with SeqCLR is the base architecture, on which we apply SeqCLR. In the paper, we used SCATTER, and here, we employ ABI-Net. In addition, here, we offer a combined stage of SeqCLR with supervised CE (this can be removed in the config).

Thank you for your reply, I am very interested in your work, I try to use seqclr to do captcha recognition, I hope to get a powerful pre-trained model.

aaberdam commented 2 years ago

Sounds interesting, good luck!