kaiqiao-rong profile photo

Kaiqiao Rong

Integrated Industrial Design MSc

I am committed to integrating machine learning with product design, exploring the innovative boundaries of smart technology and user experience. I am good at finding design inspiration through data analysis and transforming complex algorithms into simple and efficient product solutions.

BAVI – Bus Assist for Visually Impaired

According to the World Health Organization, at least 300 million people in Europe suffer from varying degrees of visual impairment. Daily travel has always been a problem for the visually impaired. Although the European public transportation system now provides many facilities and services for the visually impaired, many visually impaired people still choose to take more expensive taxis due to lack of confidence or inconvenience. This is very unfair to the visually impaired.


The goal of this project is to design a device that assists the visually impaired to access the bus, and to achieve object detection through machine learning, so that the visually impaired can access to the bus more conveniently.

How does it work?

BAVI is an assistive device dedicated to helping the visually impaired access to buses. BAVI can be easily worn on the forehead by the user and provides real-time information on bus arrivals, bus door opening and closing status, and bus door location.Users can receive real-time information from the device through Bluetooth earphones.

BAVI can be securely fastened to the forehead using an adjustable strap. The connection between BAVI and the strap is reinforced by a durable buckle, ensuring the device remains stable during use, providing reliable assistance to the user throughout their journey.
BAVI can also be attached to a cap. The connection between the BAVI and the cap is reinforced by a durable buckle that was adjusted by the designer to ensure the device remains stable and at the right angle during use.
BAVI’s outer shell is very small and light, suitable for long-term wear. The internal structure is also designed to be very simple for easy production.
We introduced three thoughtfully designed colour schemes for our product tailored specifically for visually impaired individuals. Each colour option is carefully selected to meet both functional and aesthetic needs.

Prototype testing

Through prototype testing, we went through four iterations and arrived at the final product.

During the iteration process, we improved the internal assembly problems of the product and reduced the volume and weight of the product, making the final product more convenient to carry.

Iteration 1
Iteration 2
Iteration 3
Final Prototype 1
Final Prototype 2

Computer Aided Design

Minimize the size of the product while taking assembly into consideration.
Consideration of product corners.
Final Rhino model.
Buckles designed for different wearing methods.

Tiny ML Model Training

We trained a machine learning model that can accurately identify the bus and the open and closed status of the bus door, and embedded it into the ArduCam Pico4ML-BLE.

The main selling point of this product is the use of tiny ML with low cost, low power consumption, less computing time and fast data processing.

Feature of the final model
Performance of the final model
Accuracy of the final model

Programing

Part of the code

The logic of the code is set so that the function execution logic of the device is to first identify whether there is a bus in the image. If so, then further identify the open or closed state of the bus door. If the door is open, then further identify the position of the door and inform the user.

← Previous - Charan Ravikumar
Next - Tiange Shen →