This project aims to classify waste images into two categories: Organic and Recyclable using a Convolutional Neural Network (CNN) built with TensorFlow/Keras.
The model is trained on the Waste Classification Dataset from Kaggle. It automatically learns features from images to help in proper waste segregation, which is a small step toward a cleaner and more sustainable environment.
Sourced from: Kaggle - techsash/waste-classification-data
Classes: Organic, Recyclable
Preprocessing: Resizing to (224x224), normalization, RGB conversion
A custom CNN built using:
3 Convolutional Layers + MaxPooling
Flatten → Dense Layers with ReLU activation
Dropout layers to reduce overfitting
Final Dense layer with sigmoid activation
Loss Function: binary_crossentropy
Optimizer: Adam
Epochs: 5
Accuracy visualized with matplotlib
The trained model is deployed using Gradio, offering a simple web interface to upload and classify images as:
♻️ Recyclable
🌱 Organic
Python
TensorFlow/Keras
OpenCV
Gradio
Matplotlib
Pandas, NumPy