I recently volunteered as a Machine Learning Engineer in Omdena's TrashOut project - Link.
Throughout this project, our team of 50 AI and Data Science experts from around the world worked to achieve 3 broad goals:
-
Build models on illegal dumpsites to see if there are any patterns that can help to understand what causes illegal dumpsites, predict potential dumpsites, and how to avoid such dumpsites.
-
Develop a classifier to identify if a shopping product is sustainable and environmentally friendly and to which recycling bin it belongs.
-
The project community helps to recommend how to keep the database of dumpsites updated to serve all interested parties and make cleanups more efficient.
While I contributed and collaborated in many parts of the project, one of my heaviest contributions was on experimenting with different Neural Network architectures to see if this class of AI algorithms was suitable for predicting potential illegal dumpsites.
This repository contains the jupyter notebook that I worked on with other collaborators to experiment with numerous variations of different Neural Networks. It contains explanatory documentation, descriptive graphs, and classification metrics.
Here is the official article written by the Omdena team for this project: https://omdena.com/blog/illegal-dumping
Any and all references to the code, charts, numerics, metrics, or anything else contained in this repository must be credited to Omdena with an appropriate citation.
If you would like to know or understand more about this project and how it addresses global cleanliness, read more about this project omdena.com/projects/ai-illegal-dumping.
Learn more about Omdena projects and their ideology at omdena.com.
Last but not least, if you have further questions, feel free to email me at sharmar@bxscience.edu.