This project focuses on classifying waste into various categories using Convolutional Neural Networks (CNNs), aiming to facilitate automated waste sorting and promote efficient recycling. The project uses the TrashNet dataset, a well-known dataset for waste classification. It begins with developing a basic CNN model and then enhances its performance by leveraging pre-trained models through transfer learning. Additionally, the dataset is augmented with pictures of waste items taken personally to ensure diversity and real-world applicability.
-
Data Preparation:
- Collect and preprocess the dataset, including images of waste items taken personally, resizing them, normalizing, and splitting them into training, validation, and test sets.
-
Basic CNN Model:
- Design and implement a simple CNN architecture to serve as a baseline for classification performance.
-
Transfer Learning:
- Fine-tune pre-trained models on the dataset to leverage features learned from large-scale datasets and improve classification accuracy.
-
Model Evaluation:
- Assess the performance of both the basic CNN model and the fine-tuned pre-trained models using metrics like accuracy, precision, recall, and F1-score.
-
Results and Analysis:
- Compare the performance of different models and analyze the impact of transfer learning on model accuracy and robustness.
This structure allows for a comprehensive exploration of CNN-based waste classification, from basic implementation to advanced techniques using transfer learning, incorporating real-world data.