Vision Guardian

Vision Guardian is a groundbreaking mobile application designed to enhance the independence and safety of visually impaired individuals. By harnessing the power of cutting-edge technology, the app provides:

• Comprehensive navigation support
• Real-time object detection
• Access to vital information

All of this is controlled through intuitive voice commands, making the app entirely hands-free and user-friendly.

---

Key Features

1. Voice Assistant – NOVA

The app features a voice assistant named NOVA that responds to user prompts with human-like interactions using the ChatGPT API. Key functions include:

• Speech synthesis
• Location services
• Weather updates
• Interaction with other app functions

2. Object Detection

The app uses the SSDMobileNetTensorflowLite model to detect the following objects:

Detected Objects
Airplane	Bench	Bicycle	Bird
Boat	Bus	Car	Cat
Fire Hydrant	Motorcycle	Parking Meter	Person
Stop Sign	Traffic Light	Train	Truck
Elephant	Frisbee	Horse	Sheep
Bear	Cow	Dog	Zebra
Giraffe	Skateboard	Bowl	Kite
Baseball Bat	Surfboard	Snowboard	Sports Ball
Tennis Racket	Hot Dog	Toilet	Orange
Banana	Apple	Broccoli	Carrot
Sandwich	Pizza	Donut	Cake
Microwave	Teddy Bear	Remote	TV
Hair Dryer	Wine Glass	Chair	Refrigerator
Oven	Keyboard	Mouse	Cell Phone
Bed	Vase	Couch	Dining Table

3. Obstacle Avoidance

The app helps visually impaired users avoid obstacles using an ultrasonic sensor connected via Bluetooth to an ESP32 sensor. This provides real-time feedback on nearby objects, ensuring safer navigation.

---

Hardware Setup

The hardware setup consists of:

• ESP32 Sensor: For communication via Bluetooth.
• Ultrasonic Sensor (HC-SR04): For detecting obstacles.
• Battery and TP4056 Module: To power the device.

The hardware is encased in a 3D-printed, lightweight, and durable casing, ensuring users can move freely and confidently.

---

Technology Stack

• Object Detection: SSDMobileNet TensorFlow Lite
• Voice Assistant: ChatGPT API (NOVA)
• Bluetooth Communication: ESP32 Microcontroller with Ultrasonic Sensor (HC-SR04)
• Programming Languages: Java, Python
• Platform: Android

---

How It Works

1. Obstacle Avoidance: The ultrasonic sensor detects obstacles and sends the data to the app via Bluetooth.
2. Object Detection: TensorFlow Lite identifies objects in the environment.
3. Voice Assistance: NOVA provides verbal feedback on object detection, location, and general queries.

Overall Working Flow

Working of Hardware Circuitry

---

References

1. ESP32 Documentation
2. HC-SR04 Ultrasonic Sensor Manual
3. Kaggle TensorFlow Models
4. Integrate OpenAI GPT-3.5 in Android
5. Create Personal Voice Assistant for Android
6. TensorFlow Lite Object Detection
7. TP4056 Module (YouTube)

---

Demo and Setup Videos

• https://youtu.be/ZJCVYAubN30?feature=shared
• https://youtu.be/KWKat1wZmNA?feature=shared

---

License

This project is licensed under the MIT License - see the LICENSE file for details.