BinWise AI

Home Page
WHY BinWise AI
ABOUT BinWise AI
APP Itself

Inspiration

Every day, people practice "wish-cycling"—putting non-recyclable items into the blue bin hoping they'll be recycled. This contamination often results in entire batches of good recycling being sent to landfills. I wanted to create a simple, tech-driven solution that takes the guesswork out of waste disposal.

What it does

BinWise AI is a web-based assistant that allows users to upload a photo of an item. Using a custom-trained Computer Vision model, the app identifies if the item is Plastic, Paper/Cardboard, Metal, Glass, or Trash. It then provides immediate, actionable instructions on how to properly dispose of it (e.g., "Rinse before recycling").

How we built it

As a solo developer (Team 1Bit), I built this using:

Python for the backend logic.
React for the frontend user interface.
TensorFlow/Keras to run the Deep Learning model.
Google Teachable Machine for training the initial classification model on a dataset of ~2,500 images.

Challenges we ran into

The biggest challenge was "Dataset Bias." The initial trash category was much smaller than the paper category, causing the AI to be over-confident in paper. I had to manually supplement the trash dataset with webcam images of common household waste to balance the model.

Accomplishments that we're proud of

I am proud of successfully deploying a functional AI model that can categorize waste with over 90% confidence in a solo hackathon setting.

What we learned

I deepened my understanding of image preprocessing, normalization in neural networks, and how to build a clean UI for data-driven applications using Streamlit.

What's next for BinWise AI

I plan to integrate a "Local Recycling API" to give users rules specific to their zip code and develop a real-time video detection mode for mobile devices.

Built With

google-teachable-machine
keras
numpy
pillow
python
react
tensorflow

Submitted to

Hack-Earth

Created by

As the sole member of Team 1Bit, I was responsible for the end-to-end development of BinWise AI, including:

- Data Engineering & ML: I curated and prepared a dataset of over 2,500 images across 6 waste categories. I handled the model training process using a Deep Learning approach (Keras/TensorFlow), optimizing for classification accuracy and addressing dataset imbalances in the "Trash" category.

- Backend Development: I wrote the core logic in Python to handle image preprocessing, array normalization, and model inference to ensure the AI could process user uploads in real-time.

- Frontend/UI Design: I designed and built the web interface using React, focusing on a clean, "Sustainability-first" user experience that provides immediate actionable feedback to the user.

- Environmental Research: I researched and integrated specific recycling guidelines to translate AI classifications into helpful user instructions (e.g., distinguishing between "wish-cycling" items and true recyclables).

- Project Management & DevOps: I managed the version control via GitHub, produced the technical documentation, and created the video demonstration.

Kh Isaax

Updates

Kh Isaax started this project — Feb 01, 2026 04:21 PM EST

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.