Inspiration
With problems like environmental pollution and climate-change being the first and foremost concern for most organisations, now, more than ever, there is a dire need for people to do their bit to help salvage the situation. One of the best ways to engage users and get them involved in various activities is by introducing a fun element to it, and that's where TrashIt comes in!
What it does
Once signed in to our application, users need to click pictures of empty cans/ wrappers that they'll be recycling. For every initiative that they take to recycle, users will be given some points that can be redeemed later to win some exciting deals and vouchers. In addition to this, on successfully uploading an image of the object that they'd be recycling, our proud "Trash Bin-ister" will provide them with a fact about the same.
How we built it
We went through a brainstorming session to understand what exactly we need to build out. The front end uses next.js, express.js and we have also added a layer of authentication. Typescript was used to code the backend. In addition to this, GPT-3 was used for generating random interesting facts about the object being recycled. We then trained an ML algorithm using rubbish we had accumulated while snacking during the hackathon (in other words, recycling). We also built out a deployment mechanism for custom containers on Google Cloud Platform.
Challenges we ran into
Github wasn't accessible through the wifi, which caused a lot of trouble. In addition to this, one of our teammates had to leave the venue due to their health issues.
Accomplishments that we're proud of
Even though we were behind time due to the couple of challenges that we faced, we were able to finish off our project by remotely collaborating with our teammate.
What we learned
We all stepped outside our comfort zone and picked a piece of tech stack that we weren't very familiar with and implemented it successfully.
What's next for TrashIt
Adding more objects for the object recognition feature. Building out a better UI. Training the GPT3 model better.
Built With
- auth0
- computer-vision
- express.js
- google-cloud
- gpt3
- mongodb
- next.js
- python
- pytorch
- torchserve
- typescript
Log in or sign up for Devpost to join the conversation.