Surface Defect Detection Using Deep Learning (NEU Dataset) with Streamlit App

Surface defect detection plays a critical role in modern manufacturing industries, especially in steel and metal production. Manual inspection is time-consuming, error-prone, and not scalable. In this post, I demonstrate how Deep Learning can be effectively used to detect surface defects using the NEU Surface Defect Dataset, along with a live Streamlit web application demo.

I have also recorded a full live lecture, where I explain the dataset, training process, and real-time defect prediction using a Streamlit app.

NEU Surface Defect Detection

---

📌 What This Project Covers

• Understanding the NEU Surface Defect Dataset

• Training a Deep Learning CNN model using PyTorch

• Running a Streamlit-based web application

• Uploading a test image to predict the type of surface defect

• End-to-end execution on Ubuntu 24.04

This project is highly useful for:

• Engineering students

• Research scholars

• Faculty members

• Industry professionals in manufacturing and quality inspection

---

🖥️ System Requirements

• Operating System: Ubuntu 24.04

• Python Version: Python 3.x

• Hardware:

  • CPU (training will take more time)

  • GPU (recommended for faster training)

---

⚙️ Step-by-Step Installation & Setup

1️⃣ Update the System

Open a terminal and run:

sudo apt update

---

2️⃣ Install Required System Packages

sudo apt install python3-pip python3-full git build-essential autoconf automake libxmu-dev

---

3️⃣ Create and Activate Virtual Environment

python3 -m venv ./myenv
source ./myenv/bin/activate

You should now see:

(myenv)

---

4️⃣ Install Python Dependencies

pip install notebook matplotlib torch torchvision torchaudio streamlit

⏳ This step may take some time depending on your internet speed.

---

📂 Clone the Project Repository

git clone https://github.com/tspradeepkumar/SurfaceDefectDetection.git

Navigate to the project folder:

cd SurfaceDefectDetection

---

🧠 Train the Deep Learning Model

Start the training process using:

python train_model.py

or

python3 train_model.py

⏱️ Training Time

• CPU: Takes more time

• GPU: Completes significantly faster

Once training is completed, the trained model will be saved and ready for inference.

---

🌐 Run the Streamlit Web Application

Launch the demo app using:

streamlit run app.py

This will automatically open a browser window.

---

🖼️ Using the Application

• Upload a Mono (Black & White) surface image

• The application will predict the type of surface defect

• Real-time inference using the trained Deep Learning model

📌 Please ensure the uploaded image is grayscale (monochrome) for accurate prediction.

---

🎥 Live Lecture & Demo

I have recorded a full live lecture demonstrating:

• Dataset explanation

• Model training

• Streamlit app execution

• Real-time defect prediction

👉 Follow the video for detailed explanation and live demo walkthrough.

---

🎓 Who Should Use This?

• Students learning AI / Deep Learning

• Research scholars working on Computer Vision

• Faculty conducting hands-on AI labs

• Industry professionals exploring AI for Manufacturing

---

📢 Share & Learn

If you find this project useful:

• Share it with your friends

• Recommend it to students

• Forward it to research scholars

Let’s encourage practical, hands-on learning in AI for Manufacturing 🚀