Neural Style Transfer
Combined content and style features from CNN layers using Gram matrix texture representation and a composite loss function — foundational generative vision work.
Pipeline
Architecture
End-to-end flow, stage by stage.
- 01
Content image
- 02
CNN features
- 03
Gram matrix style
- 04
Composite loss
- 05
Optimization
How it works
Implemented the image transformation pipeline in PyTorch and OpenCV, extracting content structure and artistic texture from different CNN layers.
Represented style through Gram matrices and optimized a composite objective that balances content preservation against style adoption.
Used the project to explore feature extraction, generative modeling, and loss-function design from first principles.
- Composite
- Content + style loss
- CNN
- Layer-wise feature extraction
Stack
More projects
Full-Stack
GCP + OpenAI Automation
Multi-API orchestration platform pairing GCP services with OpenAI — predictive weather, multi-participant calendar scheduling, and Gmail automation over OAuth 2.0.
Edge / IoT
Stock Prediction IoT
IoT-fed real-time market monitoring with predictive price models, multi-source feature engineering, and a Streamlit analytics dashboard.