Back to Projects
Art of Mithila - AI-Powered Cultural Heritage Classifier
Upload a photo and the model labels it as Mithila painting, Thangka, mandala, or Paubha. Next.js 15, Django, and a VGG16 model trained on those classes, with a confidence score and runner-up labels.
Project Gallery
+2
About This Project
I built a classifier for four Nepali art forms: Mithila painting, Thangka, mandala, and Paubha. You upload an image, the API preprocesses it, and a VGG16 transfer-learning model returns a label, a confidence percentage, and the next-best guesses. The UI is Next.js 15. The API is Django. Images go through Cloudinary.
Key Features
- Deep learning image classification for traditional art
- VGG16 transfer learning model implementation
- Real-time image upload and processing
- Confidence percentage display for predictions
- Alternative classification suggestions
- Cloudinary integration for image optimization
- Responsive web interface design
- RESTful API architecture
Challenges & Solutions
- Training accurate models with limited art dataset
- Implementing transfer learning with VGG16 architecture
- Optimizing image preprocessing for different art styles
- Creating intuitive user interface for complex AI predictions
- Connecting the Next.js UI to the Django API