Cleansight AI

• 

  training the model

• 

  the map that is generated for dirty locations

Inspiration

This product was created to benefit my community as I am located in an extremely polluted city and the littering issue is at its peak, most roads/public locations are completely covered in trash. This issue is being tackled by the government and a reporting system is available. This causes another issue to arise in the fact that too many areas are densely polluted and finding every location would take years. This is where my product comes into the picture, it automatically can detect when an area is polluted and pins the position on a map for easy portability. What would take years normally can be done in months using drones/public cameras using my product.

What it does

The project I have created will first determine whether a location is polluted/littered and how badly, Using that information it generates a map and pins the location allowing you to report it to the government easily. It can use pre-taken images or a drone's video/images from a drone.

How I built it

I first created a dataset of images to train the AI model on, then I found a suitable library that allows for the training of models. Lastly, I made it compatible with drone/CCTV camera footage and more efficient in the identification and creation of the map for ease of access.

Challenges I ran into

The first challenge was finding a suitable dataset to train the AI on, This problem was tackled through the use of publicly available images of my city, once this was taken care of the model was trained successfully. The next major roadblock came in the form of the fact that taking individual pictures would not work when using a drone so the next step was to make it compatible with video format. This was done by splitting the video into parts and accessing 1 frame of each part which would allow for each area to be checked while not wasting memory storing every frame.

Accomplishments that we're proud of

My main accomplishment is creating a functional AI model that consistently detects polluted areas and can allow for change in society. It makes me proud that I can actually have an influence on our planet and create a safer and more developed community for us all.

What we learned

My main learning was in the form of understanding how even a student can make a change in society and if we put our mind to something we can achieve something truly great and make a difference.

What's next for Cleansight AI

Cleansight AI is my project and it won't stop being improved, the next step would be to create an automatic report system that verifies the location and sends it to the required party for handling. Completing this would allow for many many locations to be checked and reported in turn leading to a cleaner community for us all.

Built With

• geocoder
• imageai
• python
• pytorch