Image Captioning with PyTorch – Portfolio Project

Description

This project demonstrates an Image Captioning model built with PyTorch. It uses an Encoder-Decoder architecture to generate captions for images. The project combines Computer Vision and NLP techniques and is intended as a portfolio-ready project.

The notebook is fully self-contained and allows you to:

• Train the model on your dataset
• Generate captions for any uploaded image
• Understand the workflow of a CV+NLP multimodal AI system

How to Run

1. Open notebook.ipynb in Google Colab.
2. Install the required packages (or run the first cell).
3. Execute all cells in order:
   • Setup
   • Dataset & Vocabulary
   • Encoder & Decoder
   • DataLoader
   • Training
   • Inference
4. In the "Upload Image and Test" cell, upload any image to generate a caption.

Requirements

torch torchvision nltk tqdm Pillow

Notes

• The model is lightweight and suitable for demonstration purposes.
• You can upload any image in the final cell to see the generated caption.
• This project is designed for portfolio demonstration, highlighting both CV and NLP skills.

Author

Angelo Sorte