About FoodVision
An educational project exploring the intersection of Computer Vision and modern Web Development.
Educational Journey
This project was created for educational purposes to deeply understand the mechanics of Convolutional Neural Networks (CNNs). Instead of relying solely on pre-trained models, I implemented these models from scratch using PyTorch and Torchvision.
The process involved reading original research papers, understanding the architecture, and implementing them line-by-line. I utilized Torch Transforms for data augmentation to improve model robustness. This hands-on approach provided invaluable insights into:
After implementing these models in PyTorch, I converted them to Open Neural Network Exchange (ONNX) format. This allowed me to deploy the models on the web using ONNX Runtime.
- Layer-wise feature extraction
- Backpropagation and optimization dynamics
- Overfitting mitigation strategies
- Model deployment pipelines
Frontend Architecture
- Next.js 15 (App Router) for robust server-side rendering and routing.
- Tailwind CSS v4 for modern, utility-first styling.
- Framer Motion for fluid animations and micro-interactions.
- Chart.js for visualizing prediction probabilities.
Backend Infrastructure
- FastAPI for high-performance Python API endpoints.
- ONNX Runtime for optimized model inference.
- PyTorch for model training and export.
- Python for core logic and data processing.
Built with ❤️ by CodeWizard