# Beyond Document Page Classification

We release the benchmarking code together with the proposed datasets:

* https://huggingface.co/datasets/bdpc/rvl_cdip_mp
* https://huggingface.co/datasets/bdpc/rvl_cdip_n_mp

For consistency, we add it as an anonymous model repository (can be cloned) in HuggingFace.

## Installation

The scripts require [python >= 3.8](https://www.python.org/downloads/release/python-380/) to run.
We will create a fresh virtualenvironment in which to install all required packages.
```sh
mkvirtualenv -p /usr/bin/python3 BYD
```

Using poetry and the readily defined pyproject.toml, we will install all required packages
```sh
workon BYD 
pip3 install poetry
poetry install
```

## Experiments

To replicate all experiment results from the paper, run experiments.sh

```sh
./experiments.sh
```