🏭 VisionInspect AI

AI-Powered Industrial Surface Defect Detection & Intelligent Inspection Assistant

VisionInspect AI is an intelligent industrial quality inspection system that detects surface defects using a custom PatchCore-inspired anomaly detection pipeline with ResNet50 feature extraction. The application provides anomaly visualization through heatmaps, AI-generated inspection reports using Google Gemini, downloadable PDF reports, and an interactive Streamlit dashboard for real-time quality inspection.

---

🌐 Live Demo

🔗 https://himanshu-visioninspect-ai.streamlit.app

🚀 Features

• 🔍 Surface Defect Detection using PatchCore-inspired anomaly detection
• 🧠 Deep Feature Extraction using ResNet50
• 🔥 Heatmap Visualization for anomaly localization
• 🤖 AI-generated Inspection Reports using Google Gemini
• 📄 Downloadable PDF Inspection Reports
• 📊 Confidence Score & Anomaly Score
• 🖥 Interactive Streamlit Dashboard
• ⚡ CUDA (GPU) Support
• 📁 Modular Project Structure

---

🛠 Tech Stack

Category	Technology
Language	Python
Deep Learning	PyTorch
Backbone	ResNet50
Anomaly Detection	PatchCore-inspired Method
Machine Learning	Scikit-learn
Computer Vision	OpenCV
Dashboard	Streamlit
Generative AI	Google Gemini
PDF Generation	ReportLab
Image Processing	Pillow
Numerical Computing	NumPy

---

📂 Project Structure

VisionInspect-AI/
│
├── app.py
├── train.py
├── inference.py
├── feature_extractor.py
├── patchcore.py
├── heatmap.py
├── ai_report.py
├── pdf_report.py
├── test_gemini.py
│
├── models/
├── outputs/
├── screenshots/
│
├── requirements.txt
├── README.md
├── .gitignore
└── .env

---

⚙️ System Requirements

• Python 3.11+
• Windows / Linux / macOS
• CUDA-compatible GPU (Optional)
• Google Gemini API Key

---

🚀 Installation

1️⃣ Clone Repository

git clone https://github.com/<YOUR_GITHUB_USERNAME>/VisionInspect-AI.git
cd VisionInspect-AI

---

2️⃣ Create Virtual Environment (Recommended)

python -m venv venv

Activate it:

Windows

.\venv\Scripts\Activate.ps1

Linux / macOS

source venv/bin/activate

---

3️⃣ Install Dependencies

pip install -r requirements.txt

---

4️⃣ Configure Google Gemini API

Create a .env file in the project root.

Add:

GEMINI_API_KEY=YOUR_API_KEY

Replace YOUR_API_KEY with your Google Gemini API key.

---

▶️ Running the Application

Run the Streamlit application:

python -m streamlit run app.py

Open:

http://localhost:8501

---

🧠 Training

To rebuild the Memory Bank and Threshold:

python train.py

---

🔍 Test Inference

python inference.py

---

📄 Workflow

1. Upload an industrial product image.
2. Extract deep features using ResNet50.
3. Compare extracted features with the PatchCore Memory Bank.
4. Compute the anomaly score.
5. Generate an anomaly heatmap.
6. Produce an AI inspection report using Google Gemini.
7. Download a professional PDF inspection report.

---

📷 Screenshots

Dashboard

---

Prediction & Heatmap

---

AI Inspection Report

---

PDF Inspection Report

---

📊 Project Pipeline

Product Image
      │
      ▼
ResNet50 Feature Extraction
      │
      ▼
PatchCore Memory Bank
      │
      ▼
Anomaly Detection
      │
      ▼
Heatmap Generation
      │
      ▼
Google Gemini AI Report
      │
      ▼
PDF Inspection Report

---

⚠️ Common Issues

Streamlit Not Found

Run:

python -m streamlit run app.py

instead of:

streamlit run app.py

---

Wrong Python Environment

Verify Python:

where python

Ensure it points to your active virtual environment.

---

Gemini API Quota Exceeded

If the free API quota is exhausted:

• Wait until the quota resets.
• Generate the report again.
• The application automatically shows a fallback report until the API becomes available.

---

🔮 Future Enhancements

• Multi-product support
• Batch image inspection
• Real-time camera inspection
• Database integration
• Inspection history dashboard
• Cloud deployment
• True PatchCore anomaly localization

---

👨‍💻 Author

Himanshu Singh Patel

VisionInspect AI

AI-Powered Industrial Surface Defect Detection & Intelligent Inspection Assistant

---

⭐ If you like this project

Please consider giving it a ⭐ on GitHub.