Paragraph-level-image-captioning

detail: https://www.dropbox.com/scl/fi/ybvzpnkkcy7lnn9jbkkl1/report.pdf?rlkey=grunab2372z90x0uj429x2n5o&dl=1

Environment Setup(default in download.sh)

(optional) conda create -n <name> python=3.9
pip install -r requirements.txt

Run Code From Scratch

1. Dataset loading, preprocessing and environment setup

before the following scripts, install java by yourself.

bash scripts/download.sh

2. Training(or if you only need to eval/predict, skip this step)

## tune config in scripts/train.sh
bash scripts/train.sh $output_model_path

3. Inference and evaluate

## tune config in scripts/predict.sh
bash scripts/predict.sh