https://github.com/robinmillford/art-intelligence-hub
Streamlit application that showcases advanced deep learning models for art analysis. It includes two main tools: an Artwork Authenticity classifier to distinguish between AI-generated and human-made art, and an Art Style Ensemble Classifier that identifies the artistic style of a painting.
https://github.com/robinmillford/art-intelligence-hub
ai-art binary-classification classifier grad-cam image-classification mobilenetv2 real-art streamlit-webapp transfer-learning
Last synced: about 1 month ago
JSON representation
Streamlit application that showcases advanced deep learning models for art analysis. It includes two main tools: an Artwork Authenticity classifier to distinguish between AI-generated and human-made art, and an Art Style Ensemble Classifier that identifies the artistic style of a painting.
- Host: GitHub
- URL: https://github.com/robinmillford/art-intelligence-hub
- Owner: RobinMillford
- License: agpl-3.0
- Created: 2024-11-21T03:35:48.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-08-18T09:30:45.000Z (about 2 months ago)
- Last Synced: 2025-08-18T11:27:35.191Z (about 2 months ago)
- Topics: ai-art, binary-classification, classifier, grad-cam, image-classification, mobilenetv2, real-art, streamlit-webapp, transfer-learning
- Language: Python
- Homepage: https://classify-ai-image-or-realart.streamlit.app/
- Size: 31.2 MB
- Stars: 2
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Art Intelligence Hub: An Ensemble Deep Learning Project
This project is a **college capstone project** featuring a multi-page Streamlit application that showcases advanced deep learning models for art analysis. It includes two main tools: an **Artwork Authenticity** classifier to distinguish between AI-generated and human-made art, and an **Art Style Ensemble Classifier** that identifies the artistic style of a painting.
The app is deployed on **Streamlit Cloud** for public access, enabling users to upload images and receive robust, interpretable predictions.
---
## 🔥 Features
- **Dual Classifier App**: A multi-page interface with two distinct art analysis tools.
- **Artwork Authenticity Classification**:
- Predicts whether an image is AI-generated or real art using a fine-tuned `MobileNetV2`.
- Includes **Grad-CAM heatmap visualization** to highlight the image regions that influenced the model's prediction.
- **Art Style Ensemble Classification**:
- Identifies an artwork's style from 10 categories (e.g., _Impressionism_, _Surrealism_).
- Combines the predictions of two powerful models (`EfficientNetV2-B2` and `ConvNeXt-Small`) for a more accurate and reliable result.
- **Interpretability**:
- The "Art Style" page visualizes each model's top 3 predictions in a bar chart, showing its confidence and alternative considerations.
- **Modern, Interactive UI**:
- A clean, dark-themed interface for a professional user experience.
---
## 🧠 Model Details
This project leverages three powerful, fine-tuned models:
1. **Artwork Authenticity Model**:
- **Base Model**: `MobileNetV2` (Transfer Learning)
- **Task**: Binary classification (AI Art or Real Art).
2. **Art Style Ensemble Models**:
- **Model A**: `EfficientNetV2-B2`
- **Model B**: `ConvNeXt-Small`
- **Task**: Multi-class classification (10 art styles).
- **Ensemble Strategy**: The final prediction is the averaged probability (soft voting) from both models.
---
## 📓 Kaggle Notebook
The model training and experimentation were conducted in a Kaggle Notebook. Key highlights include:
- **Dataset Preparation**:
- Combined multiple datasets, including real art and AI-generated art (`Stable Diffusion` & `Midjourney`), totaling over **150,000 images**.
- The final dataset was split into **80% training**, **10% validation**, and **10% test** sets.
- **Model Experimentation**:
- Fine-tuned three separate deep learning architectures (`MobileNetV2`, `EfficientNetV2-B2`, `ConvNeXt-Small`) using a two-stage transfer learning approach.
- Utilized modern training techniques such as the `AdamW` optimizer, learning rate scheduling, and advanced data augmentation.
- **Key Insights**:
- The ensemble approach for style classification leverages the architectural diversity of `EfficientNetV2` (a powerful CNN) and `ConvNeXt-Small` (a modern hybrid) to achieve higher accuracy than either model alone.
- `MobileNetV2` was chosen for the authenticity task due to its lightweight and efficient design, making it ideal for a fast, responsive app.
**Kaggle Notebook Link**: [Access Here](https://www.kaggle.com/code/yaminh/ai-vs-real-project)
**Kaggle Notebook Link (Art Analysis with LoRa)**: [Access Here](https://www.kaggle.com/code/yaminh/art-style-analysis-with-cnns-and-lora)
---
## 🌟 Deployment
The app is deployed on **Streamlit Cloud**, making it accessible to users anywhere.
**Streamlit App Link**: [Access Here](https://classify-ai-image-or-realart.streamlit.app/)
---
## 🖥️ How to Use
### From GitHub
1. **Fork the Repository**.
2. **Clone the Repository**:
```bash
git clone [https://github.com/RobinMillford/Art-Intelligence-Hub.git](https://github.com/RobinMillford/Art-Intelligence-Hub.git)
```
3. **Navigate to the Project Folder**:
```bash
cd Art-Intelligence-Hub
```
4. **Install Git LFS**: Your models are large files. You must have Git LFS installed to download them.
```bash
git lfs install
git lfs pull
```
5. **Install Dependencies**:
```bash
pip install -r requirements.txt
```
6. **Run the App**:
```bash
streamlit run app.py
```
7. Open the local URL (http://localhost:8501) in your browser.
### From Streamlit Cloud
1. **Visit the Deployed App**:
Open the [Streamlit App](https://classify-ai-image-or-realart.streamlit.app/) in your browser.
2. **Choose a Classifier**:
Select either "Artwork Authenticity" or "Art Style Classifier" from the sidebar.
3. **Upload an Image**:
Drag and drop an image or select a file to classify.
4. **View Results**:
See the model's prediction and explore the "Individual Model Predictions" expander for more details.
---
## 💻 How to Contribute
1. **Fork the Repository** on GitHub.
2. **Clone Your Forked Repository**.
3. **Create a New Branch**:
```bash
git checkout -b feature/your-feature-name
```
4. **Make Your Changes and Commit**.
5. **Push Changes to Your Fork**.
6. **Create a Pull Request**.
---
## 📚 Project Details
This project is part of our **college capstone project**, aimed at exploring the practical applications of **deep learning** and **computer vision** in art analysis. The goal was to develop a deployable app that combines efficient classification and intuitive user interaction.
### Technologies Used
- **TensorFlow/Keras**: For building and training the deep learning models.
- **Streamlit**: For app development and deployment.
- **OpenCV**: For image processing.
- **Git LFS**: For managing large model files.
---
## 📄 License
This project is licensed under the [AGPL-3.0 license](LICENSE).
---
## 🌟 Acknowledgments
- **Dataset Sources**: We used a combination of AI-generated art datasets and real art image collections from Kaggle.
- **Faculty Advisors**: Thanks to our professors for their invaluable guidance throughout this project.
- **Streamlit Community**: For resources and support in app deployment.
---