doctr
doctr copied to clipboard
mindee
docTR (Document Text Recognition) - a seamless, high-performing & accessible library for OCR-related tasks powered by Deep Learning.
> It's also open see for guidiance: https://mindee.github.io/doctr/v0.8.1/using_doctr/custom_models_training.html > > Only the mindee internal data which was used to train the models can't be shared because they contain sensible data...
### 🚀 The feature FAST - Detection model PyTorch: - [x] `fast_tiny` - [x] `fast_small` - [x] `fast_base` TF: - [x] `fast_tiny` - [x] `fast_small` - [x] `fast_base` Orientation models:...
### Bug description I have tried the detection model on the images which have more width than height but in all of the images, the model is not able to...
### Bug description I'm trying to run an example of doctr on my M1 Mac. Code is shown farther below. ### Code snippet to reproduce the bug ``` from doctr.io...
### Bug description I am testing the db_resnet50 on the given doctr/references/detection/evaluate_pytorch.py script. When I don't use "--rotation", the eval script is relatively fast (40s for FUNSD), but when I...
### 🚀 The feature You might consider adding COCO-text as one of the supported datasets, https://vision.cornell.edu/se3/coco-text-2/#download ### Motivation, pitch It is another high quality dataset, text on objects at various...
### 🚀 The feature - Add option to use slack for logging with `tqdm` in each training script - The logs should be clear distinguishable if you log multible runs...
### Bug description When use `sar_resnet31` I am getting this error: `expected mat1 and mat2 to have the same dtype, but got: float != c10::Half`. But some other models (I...
### 🚀 The feature TensorFlow >= 2.16.0 switches to Keras v3 (keras-core) which will especially break the saving and loading logic from models in our case - pin to >=TF...
### 🚀 The feature We should setup a github action as described in: https://huggingface.co/docs/hub/spaces-github-actions to keep the HF Space demo actual