Skip to content

Waste image classification into organic or recyclable ones with CNN algorithm.

License

Notifications You must be signed in to change notification settings

aniass/Waste-Classification

Repository files navigation

Waste Classification

General info

The project contains waste classification with Convolutional Neural Networks (CNN) algorithms to determine if it may be recycle or not. The analysis has been made in three steps: the first includes data preparation and build model CNN to waste recognise, the second includes CNN model with data augumentation and the third used a transfer learning with pre-trained MobileNet V2 model to achieve the better results.

I have created the application in Streamlit based on this trained CNN model and is available here . The code for this app is here.

Dataset

The dataset contains the 22500 images of organic and recyclable objects. It comes from Kaggle and can be find here.

Summary

The aim of the project was waste classification by using Deep Neural Networks. The dataset contains waste images recyclable and organic ones. I have built model to predict if the waste may be recyclable or not. In the analysis I have used Convolutional Neural Network (CNN) model with data augumentation and transfer learning to get more accurate predictions and choose the best one for that.

Project includes:

  • Waste classification with CNN model - Waste_cnn.ipynb
  • Waste Classification with data augmentation - Waste_Augumentation.ipynb
  • Waste Classification with transfer learning - waste_transfer_learning.ipynb

Technologies

The project is created with:

  • Python 3.8
  • libraries: TensorFlow, Keras, pillow, numpy, pandas, seaborn.

Running the project:

To run this project use Jupyter Notebook or Google Colab.