EnviroSense

Hear what's ahead and navigate with ease

Comment

Inspiration💭

According to a survey conducted by Roberto Manduchi, out of the 245 million people living with severe visual impairment, 15% hit aerial or suspended obstacles each month. Additionally, 40% of visually impaired individuals experience falls every year due to hitting such obstacles. Furthermore, almost all visually impaired people (95%) encounter obstacles while walking on the road. It's essential to note that over 40 million people worldwide are blind, and taking precautions to prevent accidents and injuries is crucial for individuals with visual impairments.

But these precautions come at steep price that many can not afford. Thus we were inspired to work on a project to save visually impaired people from such incidents and be an affordable option for all.

What it does🧭

EnviroSense is a mobile app designed to assist visually challenged individuals in detecting objects in their immediate surroundings. Using the phone's camera and advanced image recognition technology, the app can identify objects and provide audio feedback to the user. The app's user interface is designed to be user-friendly and accessible, making it easy for visually challenged individuals to navigate and use.

How we built it👷‍♂

In this project, the Flutter framework developed by Google was utilized to create a real-time object and distance detection application. Google ML Kit was used as a tool for the detection of objects in the application. The base model used in the application was TFlite, which is a lightweight machine learning framework designed to run on mobile devices.

To further enhance the accuracy of the object detection, multiple other models were incorporated into the application. These additional models were designed to work in conjunction with the TFlite model to provide more accurate and reliable object detection results.

Challenges we ran into📒

  1. Integrating AI model into Flutter SDK.
  2. Optimizing the app so it can run on previous generations of smartphone.
  3. Imroving accuracy of the AI model.
  4. Implementing narration of the detected object.

What we learned🖊️

  1. We learned to integrate an ML model to a Flutter application while keeping it light on the device.
  2. We learned about Text-to-Speech package in Flutter and its implementation across various scenarios.
  3. We learned how to support heavy AI/ML packages even for low end devices.

What's next for EnviroSense🔮

  1. Implementing the idea in a form of wearable device while keeping it affordable and accessible for all.
  2. Connecting with Government platforms and getting funding.
  3. Incorporating ML models for better visualization of data.

Built With

Share this project:

Updates