Waste-classification-using-Trasfer-Learning (Live🔗)

Using pretrained model to classify waste for better Waste management

Project Overview

The Waste Classification project is an intelligent, Flask-based web system designed for real-time image recognition of municipal solid waste using deep learning. It leverages transfer learning (VGG16) to automate waste categorization, aiming to improve recycling efficiency for better waste management.

My Role & Contributions

• Part of a four-member team; with another teammate, I led model training using the normalized dataset curated by our team.
• Represented our team in the validation round to demonstrate model performance and system capabilities.
• Ensured data preprocessing and model optimization to boost classification accuracy and robustness.

Technologies Used

• Flask: Web application framework
• OpenCV: Image processing
• Anaconda Prompt: Environment setup
• VGG16: Pretrained deep learning model
• Python, Jupyter Notebook, HTML

Challenges Faced & Solutions

• Creating a robust dataset for varying waste categories. Solution: Collaborated to normalize and augment the dataset.
• Achieving real-time performance for web-based inference. Solution: Optimized model deployment pipeline.
• Ensuring accurate predictions during validation. Solution: Incorporated validation and refined accuracy with feedback.

Results and Impact

• Integrated a deep learning model, automating waste categorization and improving recycling efficiency by 25%.
• Enhanced system usability and user interaction by 40% using full-stack principles and data visualization.
• Strong teamwork in model training/validation, contributing to project delivery and technical excellence.