environment-ai

A platform for community-contributed lightweight AI tools to help save the planet. This platform aspires to the hub for all projects related to AI and Machine Learning which encourages community engagement, especially in these times of global crisis where people feel isolated and feel helpless about solving issues around them.

Inspiration

Machine Learning and AI concepts can be intimidating for a lot of people unfamiliar with the math and science behind it. I felt I could make a few machine-learning tools online which could be accessed by anyone with a phone or a desktop browser to solve real-world problems.

What it does

It enables anyone with internet access to use machine learning tools to solve tasks related to the environment and beyond, right from their browser. It's an aspirational platform that aims to be a hub for machine learning models for environment-related tasks.

How I built it

I trained it using Tensorflow.js, a lightweight javascript library version of the more powerful C++/Python Tensorflow library. Being a javascript library it can be used with any modern browser and removes the need for local dependencies. HTML, CSS, and Javascript were to glue together the project, which also included P5.js and ml5.js libraries.

Challenges I ran into

Training using a self-collected dataset. This was done as a proof of concept exercise with results pointing towards an easy path for implementation by the public in the future.

I ran into hurdles trying to get p5.js working alongside the ml5.js libraries but finally overcame it using their docs and StackOverflow.

Accomplishments that I'm proud of

Implementing and running machine learning models on the browser resulting in a project that can be potentially scaled in the future. This can potentially be useful to people with low internet access in many developing countries.

What I learned

Different javascript libraries and technologies which enabled the implementation of the machine learning concepts.

What's next for Environment AI

• Build a production-grade platform to enable people to contribute to the growing list of machine learning tools.

• Write docs to enable people to train their environment-specific machine learning models and share them with the rest of the community.

• Find ways to make the models run offline. This could be really useful for people in developing countries with limited access to the internet.

Built With

• css3
• html5
• javascript
• ml5.js
• p5.js
• tensorflow