How to run the project code:

1. Install conda environment:
   1. cd mocha_code
   2. conda env create -f environment.yml
   3. conda activate mocha
   4. python -m spacy download en_core_web_sm
   5. python -m spacy download en_core_web_trf
2. Go to the folder "mocha_code", which is the project's folder.
3. Run the following python scripts, depending on what you want to achieve. A GPU with at least 16GB of memory is required.
   1. generate_images.py: will generate the images from the ~3200 captions
   2. generate.py: Will generate the captions with Mistral based on the available captions from GPT4
   3. generate_captions.py: Will generate captions using BLIP + MOCHa for the new dataset
   4. evaluate.py: Will evaluate the hallucination ratio using Mistral on the generated captions from generate_captions.py