🤖 Zephyrus: Robotic Arm for Surgical Assistance

Zephyrus is an innovative project integrating a robotic arm with a deep learning-based object detection system to assist in surgical procedures. This repository contains the code, documentation, and workflow details for both the hardware setup and the YOLO v11-based deep learning architecture that drives the system.

---

🔍 Overview

The project aims to overcome limitations in surgical and industrial settings by providing an affordable, reliable, and precise robotic arm solution. The system uses real-time object detection to identify surgical tools, ensuring enhanced accuracy and safety during operations.

---

📝 Problem Statement

• 💰 Cost & Reliability: Address high costs and reliability issues in surgical environments.
• 🎯 Precision: Ensure precise control mechanisms required for surgical assistance.
• 💡 Affordability: Deliver a cost-effective solution without compromising on performance.

---

🛠 Hardware Components

The project integrates several hardware components to build the robotic arm:

• Servo G90

  • 📏 Dimensions: 22.2 x 11.8 x 31 mm (approx.)
  • ⚖️ Weight: 9 g
  • 💪 Stall Torque: 1.8 kgf.cm
  • ⏱ Operating Speed: 0.1 s/60°
  • 🔌 Operating Voltage: 4.8V (~5V)
  • 🎚 Dead Band Width: 10 µs
  • 🌡 Temperature Range: 0°C – 55°C

• Servo MG996RS

  • 🔌 Operating Voltage: +5V typically
  • ⚡ Current: 2.5A (6V)
  • 💪 Stall Torque: 9.4 kgf·cm (at 4.8V); Maximum: 11 kgf·cm (6V)
  • ⏱ Operating Speed: 0.17 s/60°
  • ⚙️ Gear Type: Metal
  • 🔄 Rotation Range: 0°-180°
  • ⚖️ Weight: 55 g

• HC-05

  • 📡 Typical Sensitivity: 80 dBm
  • 🔋 Transmit Power: Up to +4 dBm RF
  • 🔌 Operating Voltage: 1.8V (with 1.8–3.6V I/O)
  • ⌨️ Default Baud Rate: 38400 (supports additional rates like 9600, 19200, 57600, 115200, 230400, 460800)

• Arduino UNO

  • 🧠 Processor: ATMega328P
  • ⏱ CPU Speed: Up to 16 MHz
  • 💾 Flash Memory: 32kB
  • 🔌 Operating Voltage: 2.7-5.5V

• MB102 PSU

  • 🔌 Input Voltage: 6.5-12 V (DC) or 5V via USB
  • ⚡ Output Voltage: Switchable between 3.3V and 5V
  • 🔋 Maximum Output Current: <700 mA
  • 🔗 Features: Onboard connectors for external devices

---

🧠 Deep Learning Workflow

The deep learning component leverages a YOLO v11-based architecture for real-time object detection. The workflow can be summarized as follows:

📂 Dataset

• Source: Labeled Surgical Tools and Images – Dataset Ninja
• Details:
  • 📸 Images: 2,620
  • 🔍 Labeled Objects: 3,639
  • 🏷 Classes: Curved Mayo Scissor, Scalpel, Straight Dissection Clamp, Straight Mayo Scissor

🧩 YOLO v11 Architecture

YOLO v11 divides the detection process into three main segments:

• Backbone: Extracts essential features from input images.
• Neck: Aggregates multi-scale features to support detection.
• Head: Performs the final object detection, outputting bounding boxes and class probabilities.

🛤 Methodology

1. Data Preparation:
   • Preprocess images and labels.
   • Split data into training and validation sets.
2. Model Training:
   • Fine-tune YOLO v11 on the surgical tools dataset.
   • Optimize hyperparameters to improve detection accuracy.
3. Evaluation:
   • Validate the model using standard object detection metrics.
4. Deployment:
   • Integrate the trained model with the robotic arm system.
   • Enable real-time detection during surgical procedures.

📅 Timeline

The project was executed in distinct phases—from hardware assembly to software integration—culminating in a live demo of the system.

---

🚀 Installation & Usage

📋 Prerequisites

• Python 3.x
• Required Python libraries (see requirements.txt), which may include:
  • TensorFlow or PyTorch
  • OpenCV
  • NumPy

⚙️ Setup

1. Clone the Repository:

   git clone https://github.com/srish-cmd/zephyrus.git
   cd zephyrus