👁️ Naya-Dristi (EchoPath) — Real-Time Spatial Assistive Pipeline

Naya-Dristi (EchoPath) is an integrated AI vision and spatial ultrasonic navigation system designed to aid visually impaired individuals. By fusing real-time YOLOv8 computer vision object tracking with high-frequency ultrasonic rangefinding, Naya-Dristi detects physical obstacles and delivers directional audio cues (LEFT, CENTER, RIGHT) via an ESP32-driven piezo sounder while maintaining a real-time HUD interface.

---

🚀 Key Features

• Multi-Modal Sensor Fusion: Combines 2D visual boundary analysis (YOLOv8) with 1D precise distance metering (HC-SR04 ultrasonic) to minimize false positives.
• Non-Blocking Firmware Engine: ESP32 state-machine firmware delivers real-time acoustic pattern pulses without interrupting sensor polling or serial message parsing.
• Directional Auditory Cues: Distinct auditory frequencies and cadence patterns indicate object direction (Left: dual-short tone; Right: single long tone; Center: rapid triple tone).
• Distance Hysteresis: Dual-threshold logic (Enter: 50 cm, Exit: 60 cm) prevents rapid alert oscillation at obstacle boundaries.
• Auto-Healing Serial Bridge: Threaded background reader in Python automatically detects ESP32 USB-Serial chips (CH340/CP210x/FTDI) and reconnects seamlessly on link drop.
• Real-Time HUD Overlay: High-visibility heads-up display built with OpenCV and Supervision showing object class, distance telemetry, alert state, and connection status.

---

🛠️ Hardware & Component Specifications

Component	Specification	Interface / Protocol	Function
Microcontroller	ESP32 NodeMCU / DevKit v1	USB-UART / PWM (LEDC)	Ultrasonic polling, buzzer pattern execution
Distance Sensor	HC-SR04 Ultrasonic Sensor	Digital I/O (TRIG/ECHO)	2 cm – 400 cm range measurement
Actuator	3.3V/5V Passive Piezo Buzzer	PWM (GPIO 4)	Directional audio feedback generation
Camera Input	External USB Webcam or Laptop Cam	USB DirectShow / V4L2	720p @ 30 FPS video feed for object detection
Host System	Windows / Linux / macOS PC	USB Serial @ 115200 Baud	YOLOv8 inference, sensor fusion, HUD rendering

---

📌 Wiring & Pinout Assignments

ESP32 Pin Connections

ESP32 Pin	Component Pin	Notes
GPIO 5	HC-SR04 TRIG	Digital Output (10 µs trigger pulse)
GPIO 18	HC-SR04 ECHO	Digital Input (Echo pulse width measurement)
GPIO 4	Piezo Buzzer (+)	LEDC PWM output (driven via transistor or direct)
GND	Sensor & Buzzer GND	Common Ground
5V / 3.3V	Sensor & Buzzer VCC	Power supply rail

---

💻 Software & Firmware Dependencies

Firmware (Arduino / PlatformIO)

• Target Core: ESP32 Arduino Core v2.x or v3.x (shim included for dual compatibility)
• Libraries: Standard ESP32 hardware drivers (HardwareSerial, esp32-hal-ledc)

Software (Python Environment)

• Python 3.10 or higher
• Core dependencies listed in requirements.txt:
  • ultralytics (YOLOv8 engine)
  • opencv-python (Image processing & HUD)
  • supervision (Annotation pipeline)
  • pyserial (USB serial bridge)
  • numpy (Tensor manipulation)

---

⚙️ Setup & Installation Guide

1. Hardware Assembly

1. Connect HC-SR04 TRIG to ESP32 GPIO 5 and ECHO to GPIO 18.
2. Connect the passive piezo buzzer positive leg to ESP32 GPIO 4 and negative leg to GND.
3. Plug the ESP32 and USB webcam into the host computer.

2. Flashing ESP32 Firmware

1. Open echopath_firmware.ino in the Arduino IDE (or VS Code with Arduino extension).
2. Select target board: ESP32 Dev Module.
3. Choose the corresponding serial port and flash the firmware at 115200 Baud.

3. Environment Setup & Execution

1. Clone the repository and navigate to the directory:

   cd herne/Naya-Dristi-main

2. Install Python dependencies:

   pip install -r requirements.txt

3. Run the pipeline launcher:
   • PowerShell:

     .\run_main.ps1

   • Direct Python:

     python main.py

---

🔊 Directional Audio Protocol

Command	Frequencies & Cadence	Meaning
LEFT	2 short beeps @ 1500 Hz	Obstacle detected on the left side of frame
RIGHT	1 long beep @ 2500 Hz	Obstacle detected on the right side of frame
CENTER	3 rapid beeps @ 2000 Hz	Obstacle directly ahead (highest priority)
NONE	Silent	Path clear / object outside threshold

---

👥 Team & Hackathon Acknowledgment

This project was built during a 1-day Team Hackathon on July 24 (from 6:00 AM to 5:00 PM).

Team Members

• Arun Bhatt
• Swastika Shrestha
• Niroj Maharjan

---

📄 License & Acknowledgments

• License: MIT Open Source License.
• Model: Powered by Ultralytics YOLOv8.
• Annotations: Built using Supervision by Roboflow.