Recyclable - If in doubt, keep it out!

If in doubt, keep it out!

Inspiration for Recyclable

We come across many instances when we don’t know if we can recycle an item and have to look it up on the web intensively to get to a conclusion. Whether it’s an used electrical appliance to a piece of broken furniture, there are ways to recycle these items; all it takes is a little research. However, not everyone has the access and enough time to research for recycling and they end up throwing away potentially recyclable materials creating a void in circular economy. But, everyone now has a smartphone with a camera, so the answer to “is this recyclable” should be made easy with the minimal possible efforts from user end, with these contexts, we have developed recyclable, an user-friendly mobile application that tells you if the item is recyclable or not after a single camera click!

How it works?

CAPTURE an item’s image, let recyclable, the app quickly identify your object and provide fast recommendations on whether the item is recyclable or not.

How we built it

Technologies used: Figma (for design), Typescript, React Native, React Navigation, Tensorflow, Tensorflowjs, OpenCV, Ascend AI, Mindspore, Mindspore XAI, Expo, Expo camera.

This is developed as a mobile application which is supported in 2 platforms - Android and ios. We trained an image classification model to classify between different materials which can either be recyclable or non recyclable. We used latest MobileNetv2 architecture which is very lite and compatible in smaller devices, This model predicts with very good accuracy at the same time it is very fast. We trained our model on both Tensorflow as well as on Ascend AI platform using Mindspore.

Challenges we ran into

Developing and deploying AscendAI model was new to us and understanding the vivid documentation needed time. However, we made sure to improve the performance in any possible way. We achieved an accuracy of 93.75 on googlenet and 91.11 accuracy with MobileNetv2. The above results depicts the power of MobileNetv2 which gave high accuracy with less parameters.

Accomplishments that we're proud of

Developing a societal good solution has a potential market with evolving technologies simplifying the long-existing problems. We are proud about our first step towards ameliorating the environment, as we all play a crucial role in the future of our planet. We have developed a platform that offers greener and sustainable solutions to the users in a comprehensive manner.

What we learned?

We learned new technologies like AscendAI, Expo, OpenCV while practicing our application development skills in React Native. We leveraged the GoogleNet network from Mindspore. We have applied different optimizers like Adam, Adam weight decay, SGD etc. We had the best experience of Using Ascend AI Platform which is very simple to use and faster computation. We also learned various business aspects to enhance the offered features.

What's next for recyclable?

We can extend the features like: COMMUNITY a platform to showcase your personalized uniquely recycled material and promote the sense of responsibility while celebrating the work. RESOURCES allowing users to access research works on recycling. BONUS POINTs for recyclers. EXPLORE where your items can be recycled based on your location. FACILITIES to see a map view of all recycling facilities close by. etc.