mindee / doctr

docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning.
https://mindee.github.io/doctr/
Apache License 2.0
3.91k stars 445 forks source link

[models] Complete the list of pretrained available models in TensorFlow and PyTorch #969

Closed felixdittrich92 closed 1 year ago

felixdittrich92 commented 2 years ago

🚀 The feature

Ensure that each model has a pretrained available checkpoint

TensorFlow:

classification:

detection:

recognition:

PyTorch:

classification:

detection:

recognition:

Motivation, pitch

felixdittrich92 commented 2 years ago

@charlesmindee i think you are the only one who can do it (data) :smiley:

nikokks commented 2 years ago

@charlesmindee, what kind of data do you use to train your text recogntions models ? Do you use custom data ? and do you use special data for french documents ?

frgfm commented 2 years ago

@nikokks for text recognition, we use cropped words from various types of documents, which sum up to about 11M images. The bias is indeed around the choice of characters (European mostly) and the language (some sequences of characters are more frequent than others).

However, a while back, we started a fully synthetic approach (cf. CharacterGenerator and WordGenerator). So far, the results are not on par, but still very decent for a fully synthetic approach :)

odulcy-mindee commented 1 year ago

Training started on Linknet_Resnet50 (Tensorflow), we'll post the results soon :wink:

odulcy-mindee commented 1 year ago

Detection models in Tensorflow are done! Now, we start recognition models in Tensorflow. First one is master!

felixdittrich92 commented 1 year ago

Closing because we track it in #1364