MD Muhaimin Rahman contact: sezan92[at]gmail[dot]com
In this project, I have tried to work on Caption generation of Images of Flickr_8k dataset. I took extensive help from Jason Brownlee's Blog article on the same dataset. But I thought some codeblocks were unnecessarily complex . So I changed them for my project. The main architecture is mainly taken from Googles paper,
Some Examples:
I have used Flickr8k dataset, which I cannot redistribute. You have to fillup this form and they will give you the dataset. You have to keep the folders Flicker8k_Dataset and Flickr_Text inside the dataset folder.
Till now, I have used features extracted from the VGG16 Model and trained on them . I think fully trainable model should improve the results which I am looking forward to work in future , God Willing.