Skip to content

I am combining esrgan and crnn to recognize objects with different approach. One of the networks based on attetion mechanizm. I am using craft to detect cracter regions. It is just an experiment.

License

Notifications You must be signed in to change notification settings

develooper1994/Superresolution_Recognition

Repository files navigation

Superresolution-Recognition

I am combining esrgan and crnn to recognize objects with different approach. It is just an experiment. Basically combining i am summing up esrgan generator loss and recognition loss. In this way i am tring to optimize both different architectures at one time.

Dataset

I am using bunch of different datasets, however my main test dataset is the UFPR-ALPR dataset.

link: http://www.inf.ufpr.br/vri/databases/UFPR-ALPR.zip

Implementation Details

Implementation Graph

  • It is just a prototype!

  • You should know that dataset class first load all dataset in to ram and starts the process. UFPR-ALPR dataset is a big one and memorizing takes very long time. read -> transform -> assign to preallocated array as an improvement.

  • Takes very long time to complete.

  • ESRGAN: it has 3 networks.

  • Ocr model has a attention and ctc loss based architecture. it has 1 networks.. Output channels order have to change.

  • There is 4 different networks have to be trained.

  • I don't have resources to continue optimization process much longer. This models are pretty heavy for any king of desktop.

  • The main deep learning framework in this repository is Pytorch

  • !!! There is not going to frequently update !!!

  • I do not provide any support or assistance for the supplied code nor we offer any other compilation/variant of it.

  • I assume no responsibility regarding the provided code.

About

I am combining esrgan and crnn to recognize objects with different approach. One of the networks based on attetion mechanizm. I am using craft to detect cracter regions. It is just an experiment.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published