Barracuda Camera Streamer

A ROS-based camera streaming system that captures video from a camera device and publishes image data for YOLO object detection inference.

Overview

This project serves as the camera input component for the barracuda-vision computer vision pipeline. It captures video frames from a camera device, converts them to ROS Image messages, and publishes them to a ROS topic for downstream processing by YOLO inference nodes.

Features

• Real-time camera capture from /dev/video0
• ROS Image message publishing to /yolo_input_image topic
• Fallback to dummy frames when camera is unavailable
• Dockerized deployment with ROS Noetic
• Configurable video resolution (640x480) and frame rate (30 FPS)

Prerequisites

• Docker and Docker Compose
• Camera device accessible at /dev/video0
• Integration with barracuda-vision repository

Quick Start

1. Build and run the containers:

   docker-compose up --build

2. The system will start two services:

   • barracuda-description: ROS core service
   • cam-to-yolo: Camera publisher service

Architecture

• ROS Node: camera_publisher publishes to /yolo_input_image topic
• Image Format: BGR8 color images at 640x480 resolution
• Publish Rate: 10 Hz
• Fallback: Black dummy frames when camera is unavailable

Integration

This streamer is designed to work with the barracuda-vision inference system. The published images on /yolo_input_image are consumed by YOLO detection nodes for real-time object detection and analysis.