The problem Eco Savvy ~ Smart Solutions for waste solves

Problem it solves - In a world burdened by inefficient waste management practices and a lack of accessible recycling information, Eco Savvy addresses the needs of individuals to guide them and gamify the waste sorting technique .

Our Key Features include :

1. SCAN THE WASTE - In the simple rhythm of data and neural networks, our deep-learning model stands as a beacon for straightforward sustainability.
   Presenting a waste segregation solution, our model harnesses the advanced ResNet50 architecture, fortified with a Softmax activation layer for heightened precision. Developed using TensorFlow and Keras, the model integrates Input, Lambda, Dense, and Flatten layers, ensuring optimal performance. Trained over 20 epochs with a dataset comprising 22,564 training images and around 2,500 test images, our approach guarantees robust learning and adaptability. This innovation is not just about waste segregation; it's a commitment to environmental sustainability.
   Upload waste image -> classify image -> gets information regarding waste material.

2. WASTE DISCOVERY BY AR ROOM - Through this we are guiding the users how to decompose any kind of waste through advanced technologies. It will help to reduce the waste generation activities by telling the users the way to reuse the waste material.

3. WASTE SORTING GAME - Our waste sorting game is an interactive and engaging experience designed to educate and challenge users in the art of proper waste disposal. User need to drag and drop the waste into particular dustbin which enhnaces their knowledge regarding waste seggregation.

4. RECYCLE HUB - This forum is made for the users so that they can post about any waste material they have
   and someone who wants to reuse them can contact the person via direct messaging.

5. CHATBOT - It uses OPEN AI API. Users can ask their quries related to waste disposal techniques/ type of waste. Our chatbot will give the users appropriate answers and solves their problem.

Challenges we ran into

Creating a strong deep-learning model for waste sorting faced many challenges. The dataset we have was small and only had two classes: organic and recyclable, causing problems with not having enough data and an uneven balance between the classes. When we were training the model, it took a long time, and the accuracy didn't turn out as good as I hoped. To improve, we checked out other projects on GitHub and used them as a reference to enhance my model. Deploying the model has brought in more challenges like compatibility issues. Even after all these efforts, the model's accuracy is around 70%. We are still working on refining it and exploring ways to make it perform better.

For AR ROOM - Cross-browser Compatibility: Different browsers and platforms interpreted A-Frame code differently, leading to inconsistencies in rendering and functionality. So, we tested the VR experience across various browsers and devices which is essential to ensure compatibility.

CHATBOT - Since it uses OPEN AI API, it gets disabled after hosting the project on github. So we solving this issue using enivronmental variables and backend services.