Thanks for stopping by! I created this project to combine my passion for AI with a real-world challenge: making cities cleaner and smarter. I wanted to see how far I could push deep learning to solve a problem that affects millions—road cleanliness.
- Inspired by the messy streets in growing cities, I wanted to automate the tedious job of road inspection.
- My goal was to build something practical—helping city planners and cleaning teams make faster, data-driven decisions.
- I challenged myself to deliver not just a model, but a deployable, user-friendly solution with real impact.
- Tackled noisy, unbalanced image data and learned to improve model generalization with transfer learning.
- Integrated multiple CNN architectures (MobileNetV2, VGG16, ResNet50) and benchmarked their strengths.
- Built an interactive web app and infographic—turning raw predictions into actionable insights for users.
- Experimented with LLM-powered strategy generation, bridging AI with urban management.
- Python for everything from preprocessing to deployment.
- TensorFlow/Keras for building and fine-tuning neural networks.
- Scikit-learn for evaluation and metrics.
- Streamlit for deploying a user-friendly web app.
- Kaggle API for seamless dataset access.
- HTML, CSS (Tailwind), JS, Chart.js for custom, interactive visuals.
- Google Gemini API for AI-driven recommendations.
- Built and compared several CNN models for binary image classification.
- Deployed the best model as a live demo: CVRoadCleanlinessClassifier
- Integrated an AI assistant to generate cleaning strategies for city teams.
- Designed an interactive infographic to make insights accessible to non-technical users.
- Expand the dataset for better coverage and robustness.
- Explore advanced architectures and real-time video analysis.
- Integrate geospatial mapping for city dashboards.
- Push edge deployment for on-device predictions.
I’m always open to feedback, collaboration, or just a chat about AI and smart cities. If this project sparks your interest, let’s connect!