Recipe and Model Creation - Overview#

This section guides the user through the complete process of creating an application recipe and the part models required for recognition and robotic picking.

Note

Prerequisites

Before proceeding with recipe and model creation, make sure that:

  • All hardware setup has been completed (Component Setup)

  • Camera calibration has been successfully completed (Camera Calibration)

  • Robot calibration is completed

  • Physical sample parts are available


Recipe vs Model: fundamental differences#

Before starting, it is important to understand the difference between a Recipe and a Model:

What is a Recipe?

What is a Model?

The global container for the entire picking application.

The specific definition of one single component to be recognized.

Includes up to 8 models, FlexiBowl parameters, Hopper settings, and communication logic.

Includes training images, ROI, visual features, filters, and robot offsets.

Manages hardware parameters such as vibrations and speed, as well as network settings such as TCP/IP port and timeout.

Manages vision parameters such as thresholds and minimum score, plus pick coordinates for the gripper.

Can handle multiple part types simultaneously in a multi-model application.

Focused on one specific visual pattern.

Tip

A recipe can contain up to 8 different models, allowing the robot to recognize and pick different part types from the same application without changing the overall configuration.


Overview of the complete process#

The process for creating a complete and working recipe is divided into several sequential stages:

Recipe and model creation workflow

Complete recipe and model creation workflow#

Main stages#

Stage

Name

Description

1

Recipe Creation

Definition of the application recipe with name, type, and FlexiBowl used

2

Physical Preparation

Positioning of the reference part in the viewing area

3

Model Training

Image acquisition and creation of the recognition pattern

4

ROI Definition

Definition of the search area where parts will be detected

5

Filter Setup

Configuration of accept threshold and recognition tolerances

6

Physical Preparation

Picking simulation with the robot to position the objects used to simulate gripper clearance

7

Coordinate Saving

Saving robot coordinates in the picking position of the reference part

8

Clearance Creation

Definition of zones that must remain free, such as gripper area or obstacles

9

Robot Coordinates

Calculation of the gripper offset for correct picking

10

Test and Validation

Verification of full operation and recipe saving



Practical advice before starting#

Material preparation#

Tip

Preparation checklist

Before starting model creation, prepare:

  • At least 10 to 20 parts of the type to be recognized for testing

  • Clean parts in good condition, representative of production

  • Gripper-clearance simulators, which must not be parts of the same type, to avoid confusion with the reference component

  • A sheet to note robot coordinates, X, Y, RZ

  • Empty and clean FlexiBowl

  • Backlight or TopLight switched on

Optimal environment#

Note

Ideal training conditions

  • Stable lighting, avoiding variable direct sunlight

  • FlexiBowl stopped

  • Robot in a safe position and not interfering during acquisition

  • FlexiVision One open with the base recipe loaded

Common mistakes to avoid#

Error

Avoid these common errors

Not saving robot coordinates during physical preparation -> impossible to calculate gripper offset

Moving the part after saving coordinates -> incorrect offset

Feature threshold too low -> model becomes too detailed and starts recognizing the background surface texture

ROI too narrow -> parts close to the edges are not detected

Clearances too small -> gripper collisions with nearby parts

Not testing with multiple parts -> issues are not discovered until production

Follow the detailed procedures in the next sections carefully to avoid these problems.


Support and additional resources#

Note

INFO buttons

For specific issues during model creation, refer to Troubleshooting.


Next steps#

Once the process overview is clear, proceed with the actual creation:

-> Start with: Create a New Recipe