The AURA Smart Specs project aims to develop an assistive technology solution for blind students, enabling them to study mathematics and navigate environments more effectively.
The system comprises wearable smart glasses with a Seeed Studio XIAO C3 microcontroller, Grove Vision AI V2 module, and various sensors, including a ToF sensor. The specs capture images of mathematical problems, process them using AI models, and provide audio feedback through a connected mobile application.
The mobile app, built with Flutter and integrated with Firebase, leverages cloud-based AI services to interpret the captured images and deliver auditory explanations. This project seeks to enhance the educational experience and independence of blind students by offering a seamless, voice-assisted interface for learning and navigation.
Build2gether2.0 ChallengeThe inspiration for this project originated from the Build2gether2.0 Inclusive Innovation Challenge, which invites participants to develop "innovative solutions to help individuals with disabilities overcome their daily struggles."
Among the challenge's solution themes, the focus on aiding visually impaired individuals particularly resonated with me, as it aligned with my interests and expertise. Having a background in designing assistive technologies and a passion for enhancing accessibility, I was motivated to tackle the theme of supporting visually impaired students in educational settings.
Problem Identification"Unless we bring in changes in the educational methods, the visually impaired will not be able to pursue education like normal people", said the World Blind Union. In the digitally inclined world of education, students with visual impairments face significant obstacles in their academic progress and learning experience. This happens due to the educational content and visual concepts delivered by the teacher are not sufficient and efficient to be grasped by the impaired students. This systemic method of teaching affects the overall educational attainment of visually impaired students creating a disadvantage for the impaired students.
"In a tragic accident, the life of a talented 24- year-visually challenged girl student of a prestigious National Rehabilitation University ended on Wednesday when she was run over by the bus of her university allegedly due to negligence of the driver in India", This was one of the crucial headlines going around with the problem unsolved. Not only does the vision affect the education but also the mobility of the students in the educational institutes, since there will be a lot of mobility needed in the institutes which affects the safety of the impaired students.
With the Huge technological advancement, the uneven opportunities can be balanced with my solution AURA, smart specs with adaptive AI voice assistant based on visual data sculpted particularly for visually impaired students for smart education and path tracking system.
What were the needs or pain points that you attended to and identified when you were solving problems faced by the Contest Masters?
- 1) DIFFICULTY IN LEARNING VISUAL CONCEPTS
- 2) MOBILITY ISSUES WITHIN UNIVERSITY
My project introduces a smart specs attachment designed to assist visually impaired students in educational settings by enhancing both their learning and mobility. The device integrates several advanced technologies, including a Grove Vision AI V2 module for real-time object and text recognition, a ToF sensor for distance measurement, an omnidirectional microphone for voice commands, and a small speaker for audio feedback. The system is controlled by a Seeed Studio XIAO ESP32-C3 board, which manages data processing and communication with a cloud-connected Flutter app. The app facilitates image analysis through Google AI services, providing audio-based study assistance. This solution aims to improve the safety and educational experience of visually impaired students by enabling them to navigate their environment and access educational materials more effectively.
Key features of the project include:- 1) Real-Time Object and Text Recognition using OCR
- 2) Audio Feedback and Voice Commands in the Flutter App
- 3) Obstacle Detection using Time-of-Flight (ToF) sensor
- 4) Cloud-connected learning using Gemini AI and Firebase
- 5) Compact and Wearable Design with easy integration into existing specs
Grove Vision AI V2 Module:
To train the machine learning model, we will utilize Google Colab, a cloud-based platform that provides powerful computational resources and allows us to run Python code in a Jupyter Notebook environment. Our dataset, sourced from Roboflow, contains labeled images specifically designed for the task of detecting mathematical equations. In the Colab environment, we will first set up the necessary dependencies and tools required for model training. Then, using the dataset from Roboflow, we will preprocess the images and annotations, ensuring they are in the correct format for training. Finally, we will initiate the training process using the YOLOv5 model, fine-tuning it on our specific dataset to optimize its performance in accurately detecting and classifying the complex equations present in the images.
The link for the Colab Notebook code can be found on my Github Repo.
Upload the Vela TFLite model to the SenseCraft AI platform by first ensuring that your TensorFlow Lite (TFLite) model is fully optimized and saved in the .tflite
format on your local system. Access the SenseCraft AI platform through your web browser and navigate to the model upload section. Once there, select the option to upload a new model, and choose the prepared TFLite file from your local storage. Follow the on-screen instructions to complete the upload, ensuring the model is properly registered and available for deployment on your AI devices.,
Function: This module is responsible for real-time object and text recognition. It captures images through its built-in camera and processes them to detect objects, text, or math equations.
Working: When the module detects an object or text, it sends this data to the Seeed Studio XIAO ESP32-C3 board via UART. It also supports saving images to an external SD card if needed.
Seeed Studio XIAO ESP32-C3 Board:
Function: Acts as the central processing unit for the smart specs. It handles communication between the Grove Vision AI V2 module, the cloud, and other components.
Working: This board processes data from the AI module, sends image data to the cloud for further processing, and receives commands from the Flutter app. It also controls the audio feedback system and manages ToF sensor and microphone interactions.
Speaker:
Function: Provides audio feedback to the user based on the processed information. The ESP32 board sends processed audio data to the speaker, which then plays back the information, such as providing navigation instructions.
Time-of-Flight (ToF) Sensor:
Function: Measures the distance between the user and objects around them. ToF sensor emits a laser pulse and measures the time it takes for the pulse to reflect. This information calculates distances and provides proximity alerts to the user.
Flutter App (Cloud-Connected):
Function: Manages data processing and user interaction through a cloud-based platform. Images and data captured by the smart specs are sent to the cloud, where they are processed using Google AI services. The processed data is then sent back to the Flutter app, which provides audio-based feedback and educational content to the user.
SD Card (for Grove Vision AI V2 Module):
Function: Stores images captured by the AI module. When an image is captured, it is saved to the SD card in a pre-defined folder. This allows for later retrieval and analysis if needed.
Power Supply (3.7V Battery):
Function: Powers all electronic components of the smart specs. Provides the necessary electrical power to the ESP32 board, AI module, microphone, speaker, and other components, ensuring they function properly.
Firebase:
Firebase serves as the backend platform to store and manage data captured by the smart specs system. The ESP32 XIAO C3, equipped with a Grove Vision AI V2 module, captures images of complex equations. These images, along with any relevant data, are then uploaded to Firebase.
Project FilesComplete project code and 3D print files are found within this project's attachments and on GitHub.
Future DevelopmentKey areas for future development include:
- Enhanced Sensor Integration: Exploring the integration of additional sensors, such as advanced proximity sensors or audio sensors, to provide more comprehensive and accurate assistance. This could improve the deviceβs ability to detect obstacles and navigate various environments more effectively
- Expanded AI Capabilities: Incorporating more sophisticated AI models and algorithms to improve the accuracy and speed of mathematical equation detection. Future updates could also include the ability to recognize and process a wider range of complex equations and mathematical symbols.
- User Feedback Integration: Continuously gathering feedback from users to refine and enhance the functionality of the device. This user-centric approach will ensure that the product evolves to meet the changing needs and preferences of visually impaired students.
In conclusion, this project demonstrates a significant advancement in assistive technology for visually impaired students. By integrating advanced machine learning models with real-time image processing and obstacle detection, the solution provides a practical and innovative approach to enhance educational experiences and personal safety. The use of cutting-edge components, including the Seeed Studio Grove Vision V2 module and the Xiao ESP32-C3 board, has been instrumental in achieving the project's goals. The successful deployment of the AI model for detecting complex mathematical equations and the seamless integration with Firebase for cloud-based image storage highlights the project's impact and potential for future developments in assistive technologies.
Thank You Note:I would like to extend our heartfelt gratitude to Seeed Studio and Build2Gether2.0 for their generous sponsorship and support throughout this project.
Comments