Geometry Attention Transformer with Position-aware LSTMs for Image Captioning

This is the official repo for the paper Geometry Attention Transformer with Position-aware LSTMs for Image Captioning.

Requirements

1. Install the dependancies in requirements.txt
2. Install java for evaluation, e.g. CIDEr, BLEU.

Train

1. train_GAT3.sh is for training our GAT model.
2. train_transformer.sh is for training the vanilla transformer model.

Evaluation

The evaluation codes are from Microsoft COCO Caption Evaluation and Consensus-based Image Description Evaluation.