Inspiration
Inspired by the fact that 70% Of the 7 to 9 billion tons of waste generated every year is mistreated, we hope to promote correct recycling practices in a creative and accessible by making trash classification easy. According to data from WHO, at least 2.2 billion people around the world are visually impaired. We might take being able to recycle for granted, but having visual impairment makes recycling a lot harder. By developing an inclusive tool for trash classification, we wish to promote recycling among the visually impaired community and providing a fresh perspective for the global waste issue.
What it does
Through this app, we strive to make trash classification easier and promote environmental consciousness in our community. By identifying trash types and offering instruction on possible disposal, our app fosters a great sense of collective responsibility. Moreover, our app benefits the visually impaired, allowing them to independently sort and dispose of their waste with the help of our app’s audio cues and OpenCV.
How we built it
We used React.js, Flask, OpenCV, Tensorflow, and ChatGPT API to construct this software.
Challenges we ran into
When we were starting to build the backend server for the program, we couldn’t decide which framework and language to use for the backend. We originally chose Node.js, but difficulties of working with a backend written in Javascript and a python script for the object detection algorithm made us choose other backend frameworks instead. We ended up choosing Flask, a python-based framework for its simplicity and shallow learning curve. It is also easier to work with Base64 formatted images passed from the frontend in Flask than in Node.js. We also found that initially the tensorflow training model achieved very high training results. However, we realized that the testing results during development are actually really low since the model was overfitting. We ended up using methods like regularization to lower the risk of overfitting.
Accomplishments that we're proud of
We managed to separate up the client-side program and the server-side program so that the client-side program will not include hardware-heavy operations. In this way the client-side program can be deployed onto various platforms that may not have the computing power to run all the algorithms. The server-side program will run on a remote machine. In the future we will move the server-side program to use cloud services such as Google Cloud or AWS. In this way we can ensure the portability of the client.
What we learned
Proficiency in using frontend framework such as react as well as back frameworks like Node.js and Flask.
What's next for BinGenius
In the future, we aim to develop more stable and more accurate object detection models for the program. Because of time constraints, the current model we use is not very accurate. If we have more time to find datasets and better models, we are sure the program will be more helpful and accurate.
Log in or sign up for Devpost to join the conversation.