Image Captioning

This project is a part of my individual projects.

-- Project Timeline: [Feb 2021]

-- Project Status: [Completed]

Project Objective

The purpose of the project is to generate captions for the given image.

Methods Used

• Deep Learning
• CNN + RNN

Technologies

• Python
• Pytorch
• Numpy, Pandas
• Jupyter
• Text Generation

Project Description

The project aims to generate text on processing the given image. Project is buit using state-of-the-art CNN and RNN deep learning models.

Dataset Description

The dataset was acquired frorm kaggle. Dataset contains

1. Images Directory containing images
2. Captions.txt file containing image_id and corresponding caption.

Steps involved to generate captions

1. Downloading dataset: Link to download dataset
2. Dataset visualization: Visualizing an image with its caption from the dataset.
   
3. Preparing custom dataset ready to feed into our CNN architecture.
4. Creating CNN + RNN architectures. Giving the output of CNN to RNN, to generate captions.
5. Training and testing our model.

Output of the test dataset

• SOS and EOS are the start of string and end of string labels.
• The model can be trained further to get better captions.