F-G Fernandez
F-G Fernandez
This issue is to be used to track the roadmap of docTR for release v0.6.0, and collect feedback from users & contributors. For the v0.5.2 roadmap, please see #967 -...
Most users of the library are more interested in existing pretrained models to use for inference rather than training. For this reason, it's important to ensure we can easily export...
Low-power devices such as Raspberry PIs are widely used by developers to come up with exciting products. Adding customized builds for ARM architectures would greatly help their efforts!
This PR introduces the following modifications: - transforms: updated the input and output signature of `RandomRotate` - character classification: expanded data augmentations for PyTorch - obj detection: switched StepLR &...
Unfortunately, one of the project dependencies does not have any conda release or any way to make one. I opened an issue on their repo https://github.com/pymupdf/PyMuPDF/issues/938 to track this, but...
The current pretrained artefact detection model was trained on a fully synthetic dataset. While this comes with several advantages, the dataset has a distribution that is still a bit far...
The library doesn't have clear information on the consequences of image transformation using different framework backends. Some need to be investigated: - appearance of artefacts during interpolation with some methods...
As discussed in #654, the artefact detection needs to improve its robustness. In order to do so and prevent overfitting, I would suggest gradually extending the list of our supported...
Add a `doctr.models.utils` module to compress existing models and improve their latency / memory load for inference purposes on CPU. Some interesting leads to investigate: - [x] FP conversion (#10)...
Currently, as specified in #609, the API template only supports single image input. With the latest version of docTR, it would be quite easy to change this to support PDF...