This repository is my solution for ai challenger Image-Captioning(Chinese).
This is based on ruotianluo's self-critical.pytorch repository.
Python 2.7 (because there is no AI_Challenger-caption-eval version for python 3) PyTorch 0.3 (along with torchvision)
You need to extract image feature for yourself, like fast-rcnn, resnet, etc.
Download ai challenger dataset and extract. First, download the ai challenger image caption data from link
Second, extract image feature.
python train.py