doctr
doctr copied to clipboard
mindee
docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning.
### 🚀 The feature This issue is to be used to track the roadmap of docTR for release v0.11.0, and collect feedback from users & contributors. - IO - Transforms...
### 🚀 The feature Dockerfile is still pinned to CUDA 11.8 now with the upgraded versions we should switch also to CUDA 12.X
- PR to keep track on Keras v3 @odulcy-mindee All models would require a retraining - maybe manual reading the weights from the keras v2 checkpoints -> set weights manual...
This PR: - Remove all TensorFlow content - Add PyTorch as fixed dependency - Drop documentation versions which rely on TensorFlow backend - Remove conda which does not support PyTorch...
### Bug description First off, thanks for the great software. Just to provide a little bit of context, we're trying to build an end-to-end system consisting of a (sort of)...
### 🚀 Adding torchscript export support are there any plans to support exporting models using torch script? ### Motivation, pitch In my experience I always felt like the loss of...
### Bug description I've modified the reference (PyTorch) training code to use `SVHN` dataset. That did not work (see traceback). By looking at what `DetectionDataset` class is doing and some...
### 🚀 The feature This issue is to be used to track the roadmap of docTR for release v1.1.0, and collect feedback from users & contributors. - IO - Transforms...