2025-ML

This repository contains the object detection and classification algorithms designed to identify and classify target objects with high accuracy and confidence.

---

Key Features

• The system utilizes three object detection models running in parallel.
• Predictions from all models are combined through a voting consensus mechanism to ensure confident results.
• Developed and trained using RoboFlow workflows

---

Architecture Overview

Process Flow

1. Input & Pre-Processing

   • Input image passed through pre-processing stages to prepare it for detection.

2. Stage 1: Detection

   • Models run independently in parallel to process the input image.
   • Each model applies a confidence threshold of 0.4 to filter out low-confidence predictions.

3. Stage 2: Voting

   • A voting system consolidates predictions from the three models.
   • Detections are validated if at least two models agree on the object.
   • The smallest bounding box is applied to the common detection for precise localization.

4. Visualization

   • Bounding boxes and labels for validated detections are generated for final output.

---

Workflow

---

Installation Instructions

Follow these steps to set up and run the project:

1. Clone the Repository

# Clone the Repository
git clone https://github.com/SchulichUAV/2025ML.git
cd 2025ML
# Install required dependencies
pip install -r requirements.txt

2. Set Up the Docker Environment

Option 1: Using Docker for Local Inference

Ensure Docker is installed and configured on your system.

# Pull the Docker image and start the container
inference server start --port 9001

Benefits of Using Docker:

• Enables offline inference
• Provides faster local detections

Option 2: Using the Roboflow API

If you prefer not to use Docker, you can run the models directly using the Roboflow API. Update the api_url in the code to: https://detect.roboflow.com

3. Run the Models

cd Scripts
python Detection.py