The supplied model is a Convolution Recurrent Neural Work trained to convert image-based subtitles into text-based
subtitles. This model was written and trained on a MacBook Pro, so the included instructions are based on a unix-like
environment. The tensorflow-metal was plugin was used for GPU acceleration. If the requirement cannot be satisfied on
your machine, use the `requirements_alt.txt` file during installation.
Requirements:
- Python 3.10.12
- Unix-like environment
Setup:
- Create a virtual environment in the root of the repository. sh```python3 -m venv .venv
2. Activate the virtual environment.
sh```source .venv/bin/activate```
3. Install dependencies (while in the activated virtual environment):
sh```(.venv) pip install -r requirements.txt
Running:
- Word-level inference model:
sh
(.venv) python3 words_inference.py
- Line-level inference model:
sh
`(.venv) python3 lines_inference.py`
The inference model runs will open an image preview through OpenCV. Advance through the images by pressing any key.