Team Name: Password1! Philip Neugebauer, Hong Sng, Tyler Wilson, Allan Nguyen AndrewIDs: pneugeba, hongsng, tylerwil, allann
We noticed that there was a lot of trash in our neighborhood and wondered if there was an efficient way to clean it all up.
What it does
Our project uses existing bus camera footage to identify areas with a high density of litter. Once our AI identifies instances of litter in a location, we plot its coordinates and density in a Google Maps heatmap.
How we built it
- We used TACO, an open image dataset of waste in the wild. It contains photos of litter taken under diverse environments, from tropical beaches to London streets. These images are manually labeled and segmented according to a hierarchical taxonomy to train and evaluate object detection algorithms.
- We then used Detectron 2, an Open-Source PyTorch-based object detection library. We employed its pretrained region-based convolutional neural network and retrained the model using TACO dataset to detect litter.
Challenges we ran into
- It was difficult to install detectron2 on Windows (it was not officially supported for Windows)
- Issues uploading huge data sets to Google Drive and Collaboratory
- 3/4 members were having a hackathon for the first time
- Obtaining GPS data from photos
Accomplishments that we're proud of
- We managed to successfully train our AI to detect litter in a short time frame
- We created a functional solution to our problem that can be used in other situations by other people
- Using a heatmap, we created an easily understandable visual representation of litter density
- We all started as strangers but we made a great team :)
What we learned
- Our first time retraining a model on new data
- Experience with Detectron2, Google Collaboratory, Google Maps API
What's next for AI Trash Detection Heatmap
- We will publish our project as open-source on GitHub and add complete instructions on how to install and use it so that others can benefit from it
- We are interested in possible applications of our project with drones or robots, e.g. Drones or robots picking up trash with our AI.