This is a repository for the final project in Deep Learning School about NLP spring 2021. Here I try to solve image captioning problem.

The whole idea of image captioning is to create the description of one picture. In two architectures below I am using the method of first giving the picture to the model and then making generation according to it and the previously decoded state.

Creating this project I've used three articles: Overview of image captioning models Show, Attend and Tell Metrics for image captionin

In a few scipts you can find:

• model - with and without attention mechanism
• training pipeline
• metrics calculation

All of the parts of the models and its representation can be found here in image_captioning_project.ipynb

All of the training reports with metrics can be found here:

model without attention

model with attention