🏭 Casting Defect Detection using CNN

A deep learning solution built to classify defective vs non-defective metal castings.
This project uses a Convolutional Neural Network (CNN) and is deployed using Gradio + Hugging Face Spaces for real-time inference.

---

📌 Table of Contents

• Overview
• Features
• Dataset
• Model Architecture
• Training
• Results
• Demo
• How to Use
• Project Structure
• Installation
• Author
• Connect With Me

---

📘 Overview

Metal casting defects can cause massive losses in manufacturing industries.
This project automates defect detection using a trained CNN model that classifies images into:

• Defective
• OK (Non-defective)

The model is optimized, saved as best_model.h5, and served through a Gradio interface.

---

✨ Features

✔️ Real-time image classification
✔️ Clean, simple Gradio UI
✔️ Transfer-learning–ready architecture
✔️ Works on CPU (no GPU required)
✔️ Fully deployable on Hugging Face Spaces

---

📂 Dataset

Casting Defect Dataset containing:

Class	Count
Defective	~3,900
OK (Non-defective)	~3,900

Dataset Structure:

casting_data/
 ├── train/
 │    ├── defective/
 │    └── ok/
 └── test/
      ├── defective/
      └── ok/

---

🧠 Model Architecture

The model is a CNN trained from scratch:

• Conv2D + ReLU
• MaxPooling
• Dropout
• Fully Connected Layers
• Output Softmax Layer

Loss: binary_crossentropy
Optimizer: Adam
Metrics: accuracy

---

📊 Training

• Image Size: 224×224
• Epochs: 25–30
• Augmentation: Yes
• Best model saved as: best_model.h5

---

✅ Results

Metric	Score
Training Accuracy	~98%
Validation Accuracy	~97%
Test Accuracy	~97%

---

🚀 Demo

You can try the live model here:

🔗 Hugging Face Demo: (https://huggingface.co/spaces/RayanAIX/casting-defect-detector)

📷 Screenshot

---

🛠 How to Use Locally

pip install -r requirements.txt
python app.py

---

📁 Project Structure

├── app.py                 # Gradio application
├── requirements.txt       # Dependencies
├── best_model.h5          # Saved model
├── dataset/               # Optional dataset folder
├── README.md

---

⚙️ Installation

git clone https://github.com/<your-username>/Casting-Defect-Detection-CNN.git
cd Casting-Defect-Detection-CNN
pip install -r requirements.txt
python app.py

---

👤 Author

Muhammad Rayan Shahid

---

🔗 Connect With Me

Platform	Link
GitHub	https://github.com/RayanAIX/
LinkedIn	https://www.linkedin.com/in/muhammadrayanshahid/
Kaggle	https://www.kaggle.com/muhammadrayanshahid
Hugging Face	https://huggingface.co/RayanAIX
YouTube	https://youtube.com/@ByteBrillianceAI

---

⭐ If you like this project, give it a star!