cvfe 0.2.5

Canada Visa Forms (5257e and 5645e) Extractor.

Canada Visa Form Extraction

Canada Visa Form Extraction (CVFE) is a tool set to extract and transform Canada visa forms (IMM 5257 E and IMM 5645 E) into standard format.

1 Installation

1.1 Package

1.1.1 PIP

pip install git+https://github.com/Nikronic/canada-visa-form-extraction.git

Or if you have cloned/downloaded the repository already, just use pip install .

1.1.2 Using Conda Environment File

We have provided a yml file (conda_env.yml) for ease of install using conda (or mamba). Please use:

conda env create -n YOUR_ENV_NAME --file conda_env.yml

1.1.3 Using Docker

Official Image: The easiest way is to just pull the official image from GitHub Container Registry:

docker pull ghcr.io/nikronic/cvfe:DESIRED_VERSION

For instance, for the version v0.2.1 (cat VERSION), you can pull docker pull ghcr.io/nikronic/cvfe:v0.2.1.

Then for running it, use:

docker run -p YOUR_HOST_PORT:CONTAINER_PORT ghcr.io/nikronic/cvfe:DESIRED_VERSION python -m cvfe.main --bind 0.0.0.0 --port CONTAINER_PORT

For example:

docker run -p 9999:8000 ghcr.io/nikronic/cvfe:v0.2.1 python -m cvfe.main --bind 0.0.0.0 --port 8000

note: All the images can be found on the repo packages.

Manual Image: We have provided a Dockerfile for ease of install using docker. Please use:

docker build -t cvfe .

For running it, you need to map the port 8000 of the container to your desired port (8001) on host:

docker run -p 8001:8000 cvfe

remark: note that we do not use mamba in Dockerfile as the complexity of the environment is small and the overhead of installing mamba itself might not worth it. Nonetheless, we never tested this hypothesis and appreciate your feedback.

1.2 Manual Installation

Using conda as the package manager is not necessary nor provides any advantages (of course virtual environment are necessary). Hence, if you like, you can install dependencies via conda. Note that only few of the dependencies are available on conda channels and you still need to use pip.

From now on, the phrase "if conda" assumes if you are using conda for your packages. Otherwise, pip only is inferred.

tip: You can use mamba to hugely speed up the installation process. If you don't want to, replace all instances of the mamba with conda in following steps.

[Optional] 1.2.1 Create a conda env

if conda: conda create --name cvfe python=3.11.0 -y

[Optional] 1.2.2 Activate the new environment

Make sure you activate this environment right away:

if conda: conda activate cvfe

1.2.3 Update pip

You should have at least pip >= 23.1.2

pip install --upgrade pip

[Optional] 1.2.4 Pin the Python version

When using conda, mamba and so on, it might update the Python to its latest version. We should prevent that by pinning the Python version in the conda environment. To do so:

if conda: echo "python 3.11.0" >>$CONDA_PREFIX/conda-meta/pinned

1.2.5 Install Processing Dependencies

We need pandas and numpy and some others, but installing pandas will install all those needed.

pip install pandas if conda: mamba install pandas

1.2.5 Install data extraction dependencies

1. pip install xmltodict>=0.13.0. if conda: mamba install -c conda-forge xmltodict>=0.13.0 -y
2. pip install pikepdf>=5.1.5
3. pip install pypdf2>=2.2.1

1.2.6 Install API libs

These libraries (the main one is FastAPI) are not for the ML part and only are here to provide the API and web services.

1. pip install pydantic>=1.9.1
2. pip install fastapi>=0.85.0
3. pip install gunicorn>=20.1.0
4. pip install uvicorn>=0.18.2
5. pip install python-multipart>=0.0.5

[Optional] For making it online (only for development) using ngrok:

1. pip install pydantic-settings
2. pip install pyngrok

For using ngrok, start uvicorn server with your own args:

USE_NGROK=True python api.py --bind host --port port

Note that USE_NGROK=True has been handled by the code and you can use this flag to use ngrok (online) or not (offline). Also, you can use 0.0.0.0 for the host to listen on all interfaces. By default we use host=0.0.0.0 and port=8000.

---

1.2.7 Install this package cvfe

Make sure you are in the root of the project, i.e. the same directory as the repo name. If you are in correct path, you should see setup.py containing information about cvfe.

pip install -e .