/NIC

Noise-Aware Image Captioning with Progressively Exploring Mismatched Words

Primary LanguagePythonMIT LicenseMIT

Noise-Aware Image Captioning with Progressively Exploring Mismatched Words

The code for deploying the NIC framework on AoANet.

Requirements

  • Python 3.6
  • Java 1.8.0
  • PyTorch 1.7.0
  • cider (link)
  • coco-caption (link)
  • tensorboardX

Training

Prepare data

You should preprocess the dataset and get the cache for calculating cider score for SCST:

$ python scripts/prepro_ngrams.py --input_json data/dataset_coco.json --dict_json data/cocotalk.json --output_pkl data/coco-train --split train

Start training

$ sh train.sh

Acknowledgements

This repository is based on AoANet , and you may refer to it for more details about the code.