🥇 RecycleAid is one of the Top Performing Teams of HackTable 2021
Team
- Thalen Abadia
- Gourav Khounger
- Erin McGurk
- Aldrich Alvarez
Inspiration
Our team was looking at ways one can solve global climate issues. While thinking for ways to address this, we realized that wastage of reusable things is a common thing that we can see throughout the world. We all know that things must be recycled, but many times we don't bother knowing the correct way to do so.
What it does
RecycleAid is a web app where one can scan an image to look for recyclable material and the way to recycle them. The app also shows images of different creative ways one can repurpose the recyclable material! A lot of people around the world do not know how to recycle properly, and that is causing big issues with waste. This app can teach and help those who don't know how to recycle, and it can help reduce waste. When so much of the world isn't recycling properly, there is a big impact on land, water, and air to a large extent, by either causing pollution or harming aquatic/terrestrial life. Our app promotes the conservation of the environment by teaching people how they can recycle different kinds of waste and materials which will help contribute to the SDGs.
RecycleAid is a web app where one can scan an image to look for recyclable material and the way to recycle them. The app also shows images of different creative ways one can repurpose the recyclable material! A lot of people around the world do not know how to recycle properly, and that is causing big issues with waste. This app can teach and help those who don't know how to recycle, and it can help reduce waste. When so much of the world isn't recycling properly, there is a big impact on land, water, and air to a large extent, by either causing pollution or harming aquatic/terrestrial life. Our app promotes the conservation of the environment by teaching people how they can recycle different kinds of waste and materials which will help contribute to the SDGs.
How we built it
The web app is built using anvil- which allows one to create python only web apps easily, without needing to format or do as much coding on the fronted part. It is a valuable tool we found during our research on the best way we could connect our AI model, trained in Jupyter Notebook using python, with our web app. Anvil provides a really easy way to do so. We've used HTML and CSS to design the website, embedding the anvil app in the site, and then we used Github pages to host the website. We used a garbage classification data-set from Kaggle to train our model, and using PyTorch we were able to create the image classification.
The web app is built using anvil- which allows one to create python only web apps easily, without needing to format or do as much coding on the fronted part. It is a valuable tool we found during our research on the best way we could connect our AI model, trained in Jupyter Notebook using python, with our web app. Anvil provides a really easy way to do so. We've used HTML and CSS to design the website, embedding the anvil app in the site, and then we used Github pages to host the website. We used a garbage classification data-set from Kaggle to train our model, and using PyTorch we were able to create the image classification.
Challenges we ran into
The major issue we faced during the development is getting the image upload to work. The anvil app was able to get the file URI from the local machine but it was not getting saved to the development server in the Jupyter Notebook directory. We tried different ways like converting the image to base64 format and then sending it through anvil server callable functions, but to no avail. But, at the end, we managed things by writing the file data to a new file and saving that and it worked! Putting all the pieces together from the data-set, to Jupyter, to Anvil, and then finally to Github proved to be a challenge but we were able to put it all together in the end.
The major issue we faced during the development is getting the image upload to work. The anvil app was able to get the file URI from the local machine but it was not getting saved to the development server in the Jupyter Notebook directory. We tried different ways like converting the image to base64 format and then sending it through anvil server callable functions, but to no avail. But, at the end, we managed things by writing the file data to a new file and saving that and it worked! Putting all the pieces together from the data-set, to Jupyter, to Anvil, and then finally to Github proved to be a challenge but we were able to put it all together in the end.
Accomplishments that we're proud of
We are proud of our project because of our inexperience with machine learning. Most of us had either barely used or never used machine learning before, and we were still able to create a functioning machine learning program that can benefit the world. We believe that this web app could be used in the real world and could be very helpful to users and to the planet, and we are proud of that.
We are proud of our project because of our inexperience with machine learning. Most of us had either barely used or never used machine learning before, and we were still able to create a functioning machine learning program that can benefit the world. We believe that this web app could be used in the real world and could be very helpful to users and to the planet, and we are proud of that.
What we learned
We learning a lot about machine learning. Most of us were inexperienced in machine learning and we were able to pick it up quickly. We also learned how to use Anvil and Jupyter Notebook well, and that helped us learn how python can be connected to web apps.
We learning a lot about machine learning. Most of us were inexperienced in machine learning and we were able to pick it up quickly. We also learned how to use Anvil and Jupyter Notebook well, and that helped us learn how python can be connected to web apps.
What's next for RecycleAid
We are very happy with how RecycleAid turned out, but if we had more time we would have gotten a larger data set so that the program could be more accurate. Predicting materials from an image is no easy task for a computer, but we could work to improve it and train the data more. We could also create an Android app so that mobile users can have an even better experience.
Github
We are very happy with how RecycleAid turned out, but if we had more time we would have gotten a larger data set so that the program could be more accurate. Predicting materials from an image is no easy task for a computer, but we could work to improve it and train the data more. We could also create an Android app so that mobile users can have an even better experience.
Github
