✈️ Aircraft Damage Classification (Crack vs Dent) & 🖼️ Image Caption Generation

This repository contains my Final Project for the course:
“Introduction to Deep Learning & Neural Networks with Keras”

The project has two major components:

1. Aircraft Damage Classification – A Convolutional Neural Network (CNN) with Transfer Learning (VGG16) to classify aircraft damages into Cracks vs. Dents.
2. Image Caption Generation – Using BLIP (Bootstrapping Language-Image Pretraining) to generate meaningful captions for unseen images.

---

📌 Project Overview

• ✅ Built and trained a CNN for aircraft damage classification with >81% test accuracy.
• ✅ Classified type of damage (Crack / Dent) instead of just binary damaged vs. normal.
• ✅ Applied Transfer Learning (VGG16) with data augmentation and callbacks (EarlyStopping, ReduceLROnPlateau).
• ✅ Implemented BLIP for image captioning that generates human-like captions.
• ✅ Explored both computer vision (classification) and vision-language (captioning) tasks in one project.

---

🛠️ Tech Stack

• Frameworks: TensorFlow, Keras, PyTorch
• Models: CNN, VGG16 (Transfer Learning), BLIP
• Languages: Python
• Tools: Jupyter Notebook

---

📂 Project Structure

  
├──aircraft_damage_dataset_v1/
|  ├── train/
|  │   ├── dent/
|  │   └── crack/
|  ├── valid/
|  │   ├── dent/
|  │   └── crack/
|  └── test/
|      ├── dent/
|      └── crack/ 
├── aircraft_damage_dataset_v1.tar
├── Damage_Detection_and_BLIP_Captioning # BLIP caption generation
├── download_dataset.py                  # Download the dataset
├── requirements.txt                     # Required dependencies
├── .gitignore                           # Avoid venv and dataset
└── README.md                            # Project documentation
  

---

⚡ How to Run

1. Clone this repo:

   git clone https://github.com/buildwithmehul/Aircraft-Damage-Detection-and-BLIP-Captioning.git
   cd Aircraft-Damage-Detection-and-BLIP-Captioning
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Download the dataset:

   python download_dataset.py
   

4. Run the Jupyter notebook:

• Damage_Detection_and_BLIP_Captioning.ipynb

---

🚀 Results

Aircraft Damage Classification (Crack vs Dent)

• Training Accuracy: ~85%
• Test Accuracy: ~81%
• Model Used: VGG16 (frozen base + fine-tuned dense layers)

📊 Example:

Image	Prediction
	Crack
	Dent

Image Caption Generation (BLIP)

BLIP generates captions for unseen images:

• Input: 🖼️ Aircraft flying in cloudy sky

• Output: "an airplane flying through the clouds"

• Input: 🖼️ Damaged aircraft

• Output: "a damaged airplane parked on the runway"

---

📜 Certificate

This project was submitted as part of my completion for Introduction to Deep Learning & Neural Networks with Keras (IBM).
🎓 Certificate available here → Introduction to Deep Learning & Neural Networks with Keras

---

🙌 Acknowledgements

• IBM Skills Network : Introduction to Deep Learning & Neural Networks with Keras
• BLIP (Bootstrapping Language-Image Pretraining) by Salesforce
• Aircraft damage dataset (publicly available)

---

👤 Author

Mehul Khanna 🔗 Linkedin 🔗 Github

---

📄 License

This project is licensed under the MIT License. — see the LICENSE file for details.