Image Captioning

Overview

• Image captioning model which predicts the caption of a given image.
• The dataset used is COCO 2014.
• Faiss Library (https://faiss.ai/) is used for efficient similarity search and clustering.

Working

• Image captioning works by using KNN to find the 5 closest image to our given image and then picking from the image caption which is most similar to our image.
• For getting 5 closest neighbours, we use Faiss library which is an efficient way for similarity searching and clustering.

Execution

• Run this notebook in jupyter and run all cells.
• It will also download the COCO dataset.

Note

• The notebook accesses the COCO dataset through mounted google drive, change it to point to where the COCO dataset is available.