Overview

Time: ~55 minutes

Scenario

This tutorial uses the simplest work cell (camera + compute) to teach patterns that apply to all Viam applications.

You’re building a quality inspection station for a canning line. Cans move past a camera on a conveyor belt. Your system must:

1. Detect when a can is present
2. Classify it as PASS or FAIL (identifying dented cans)
3. Log results for review and analysis
4. Provide a monitoring dashboard for operators

Tutorial

In this tutorial you will work through a series of tasks that are common to many robotics applications. The techniques you learn here are applicable regardless of what hardware, software, data, or machine learning models you are working with.

Full section outline

Gazebo Simulation Setup (~10 min)

• Prerequisites
• Step 1: Build the Docker Image
• Step 2: Create a Machine in Viam
• Step 3: Create a credentials file
• Step 4: Start the Container
• Step 5: Verify the Setup

Part 1: Vision pipeline (~10 min)

• 1.1 Verify your machine is online
• 1.2 Locate your machine part
• 1.3 Configure the camera
• 1.4 Test the camera
• 1.5 Add a vision pipeline with a fragment

Part 2: Data capture (~5 min)

• 2.1 Configure data capture
• 2.2 View captured data
• 2.3 Summary

Part 3: Control logic (~10 min)

• 3.1 Generate the module scaffolding
• 3.2 Add remote machine connection
• 3.3 Add detection logic
• 3.4 Summary

Part 4: Deploy a module (~10 min)

• 4.1 Review the generated module structure
• 4.2 Build and upload your module
• 4.3 Add the module to your machine
• 4.4 Configure detection data capture
• 4.5 Summary

Part 5: Productize (~10 min)

• 5.1 Create a workspace
• 5.2 Add a camera stream widget
• 5.3 Add a defects per minute widget
• 5.4 Add a confidence trend widget
• 5.5 Arrange your dashboard
• 5.6 Summary

Get started

Before starting, set up the Gazebo simulation environment by following the Gazebo Simulation Setup guide (~10 min).