👁️ NeuroVision

Cognitive Visual Attention Simulation using Saliency and Fixation Modeling

A cognitive AI simulation that mimics how humans visually explore a scene using saliency maps and fixation-based attention scanning. Upload an image and watch how a cognitive system prioritizes visual information.

---

📖 What is NeuroVision?

NeuroVision is a Cognitive AI simulation system that demonstrates how human visual attention works when observing a scene.

Instead of processing images uniformly, humans naturally focus on regions of interest. This project simulates that behavior using saliency maps and fixation-based scanning, allowing users to observe how an artificial cognitive system prioritizes visual information.

The system works as follows:

1. A user uploads an image representing a visual scene.
2. The system generates a saliency map highlighting visually prominent regions.
3. The algorithm simulates fixation points, mimicking how human eyes move across a scene.
4. The results are visualized interactively through a Streamlit interface.

This provides a simple yet powerful demonstration of attention modeling in cognitive AI systems.

---

✨ Features

Feature	Description
📸 Image Upload Interface	Upload any PNG or JPG image for simulation
🎯 Saliency Map Generation	Highlights visually prominent regions
👀 Fixation Simulation	Simulates eye movement across salient regions
⚙️ Adjustable Parameters	Customize blur kernel size, threshold, and fixation count
📊 Interactive Visualization	Real-time UI using Streamlit
🧠 Cognitive AI Concepts	Demonstrates perception and attention modeling
🎓 Educational Tool	Ideal for AI and Cognitive Science coursework

---

🏗️ System Architecture

Visual Attention Pipeline

graph TD
    A[👤 User Uploads Image] --> B[Streamlit Web Interface]
    B --> C[Image Preprocessing]

    C --> D[OpenCV Saliency Detection]
    D --> E[Saliency Map Generation]

    E --> F[Threshold Filtering]
    F --> G[Top Salient Regions Selection]

    G --> H[Fixation Simulation]
    H --> I[Fixation Points Overlay]

    I --> J[Visualization Output]

Loading

---

Cognitive Processing Workflow

The simulation replicates a simplified version of human visual attention:

1️⃣ Scene Perception

• The uploaded image acts as a visual stimulus.

2️⃣ Saliency Computation

• OpenCV detects regions that visually stand out based on:

  • color contrast
  • edge intensity
  • brightness variations

3️⃣ Attention Filtering

• A threshold filter removes low-importance regions.

4️⃣ Fixation Selection

• The system selects top N salient points.
• These points simulate human eye fixation behavior.

5️⃣ Visualization

• Displays:

  • Original image
  • Saliency heatmap
  • Fixation points overlay

---

🛠️ Technology Stack

Core System

Component	Technology
Programming Language	Python 3.9+
Web Interface	Streamlit
Computer Vision	OpenCV
Numerical Processing	NumPy
Visualization	Matplotlib

---

📂 Project Structure

visual_attention_streamlit/
│
├── app.py              # Main Streamlit application
├── attention_model.py  # Saliency and fixation logic
├── utils/
│   └── image_utils.py  # Image preprocessing utilities
│
├── requirements.txt    # Project dependencies
└── README.md           # Project documentation

---

🚀 Installation & Setup

Prerequisites

• Python 3.9+
• pip

---

1️⃣ Clone the Repository

git clone https://github.com/your-username/visual_attention_streamlit.git
cd visual_attention_streamlit

---

2️⃣ Create a Virtual Environment (Recommended)

python -m venv venv

Activate Environment

Mac/Linux

source venv/bin/activate

Windows

venv\Scripts\activate

---

3️⃣ Install Dependencies

pip install -r requirements.txt

---

🏃 Running the Application

Launch the Streamlit application:

streamlit run app.py

The application will start locally at:

http://localhost:8501

---

🌐 Application Workflow

Step 1 — Upload Image

Upload a PNG or JPG image representing the visual scene.

---

Step 2 — Adjust Simulation Parameters

Using the sidebar controls, configure:

• Blur kernel size
• Saliency threshold
• Maximum fixation points

---

Step 3 — Generate Attention Simulation

The system processes the image and generates:

• Saliency heatmap
• Fixation simulation

---

Step 4 — Analyze Output

The UI displays:

Output	Description
Original Image	Input visual scene
Saliency Map	Regions attracting attention
Fixation Points	Simulated eye movement

---

🧠 Cognitive AI Concepts Used

Concept	Description
Saliency Map	Highlights visually important areas
Fixation	A point where the eye briefly focuses
Visual Attention	Prioritization of relevant visual stimuli
Thresholding	Filters out less important visual data
Cognitive AI	AI inspired by human perception and cognition

---

🐛 Known Issues & Troubleshooting

Saliency detection not working properly

Ensure:

• Images are not extremely low resolution
• Contrast between objects is sufficient

---

Streamlit UI not loading

Verify the server started successfully:

http://localhost:8501

Restart the app if needed.

---

🔮 Future Improvements

• 👁️ Real eye-tracking dataset integration
• 🎥 Video attention simulation
• 📊 Temporal attention heatmaps
• 🤖 Deep learning attention models
• 🧠 Object detection + attention fusion

---

🎓 Academic Value

This project demonstrates key concepts from Cognitive Artificial Intelligence, including:

• perception modeling
• visual attention mechanisms
• cognitive psychology principles
• computational vision techniques

It serves as a lightweight educational simulation for understanding how cognitive systems prioritize information.

---

👨‍💻 Author

Kishore P CSE (AI & Robotics) – VIT Chennai

---

📜 License

This project is licensed under the MIT License.

---

Understanding how machines can perceive the world like humans.

NeuroVision — modeling attention in artificial cognitive systems.