本文原版参考Harshall Lamba的作品实现 博客地址:https://towardsdatascience.com/image-captioning-with-keras-teaching-computers-to-describe-pictures-c88a46a311b8 代码地址:https://github.com/hlamba28/Automatic-Image-Captioning 打印keras模型,需要安装graphviz:apt-get install graphviz;需要安装pip install pydot; 本文用到glove词嵌入模型参数:https://nlp.stanford.edu/projects/glove/ to do: 1、增加Flicker30k进行测试 2、Image嵌入层和caption嵌入层采用拼接方式 3、词汇表词频,如何处理pad_sequence中填充问题 4、对比im2txt方法
leena201818/ImageCaption
add captions to images of Filcker8k dataset using keras,reference Harshall Lamba's work.
Python