Automatic Tag Detection and Classification with Cintoo AI

This article describes the AI tag detection algorithm and outlines categories of assets that could be detected.

Watch tutorial video: AI Detection - Geometry

Permissions: users with roles including View Tags and Manage Tags permissions can launch AI tag detection.

TABLE OF CONTENTS

Algorithm
Categories of Objects Detected
How to use

Algorithm

Cintoo's AI Engine replaces the cumbersome manual tag search within scan data with an automated, faster, easier, and more accurate approach.

Based on a Deep Neural Network engine and modern computer vision algorithms, Cintoo's AI Engine allows users to automatically geolocate and classify tags within 3D scans. Cintoo's AI Engine is tailored to industrial customers’ requirements and important objects they need to identify for their specific applications.

For example, Cintoo's AI Engine can readily detect common Industry 4.0 objects such as valves, flanges, electric motors, electric devices etc. Cintoo AI Library contains a large set of use-case content that has been manually annotated. This customized list of assets is continuously enhanced and updated according to customer requirements.

Note: unlimited tagging capacity is available only with TWIN edition, while BIM edition users do not have access to this functionality.

Learn more about editions BIM and TWIN.

Categories of Objects Detected

Equipment

AISBOXes

Control boxes

DC Controllers

Electric boxes

Electric motors

Fans

Filters

Flanges
Gas Flow Meters
Pipe Trees

Pressure gauges
Pressure regulators

Safety Release Valves
Slam Shut Valves
Speakers

Surveillance cameras

Train Warning Signs
Valves (Hand Wheel Actuator)

Valves (Handle Actuator)

Fire Signage and Safety Equipment

Alarm Activators

Detectors

Emergency Exit Signs

Emergency Shut Down Buttons

Extinguishers

Hoses

Safety Signs

Sounders

Sprinklers

Geometry

Pipes/Cylinders - detects any object of cylindrical form including pipes
Planes - detects any plane surfaces like walls, doors etc.

How to use

1. In 3D View, select a work zone that contains scans to use for AI detection process.

Note: current AI Engine will search for tags in a maximum of 500 scans simultaneously, so please organize scans accordingly.

2. Using Tag Explorer, select a tag list that has been created for this search. If no lists are present see Create a List of Tags.

3. With a tag list selected, click AI Detection in the list of available actions in Tag Explorer.

4. Select tag class(es) to be automatically detected and click Detect.

All tags are divided by categories for better distinction between them. Now there are three categories Equipment, Fire Signage and Safety Equipment and Geometry. More categories would be introduced with new releases.

Online detection process benefits from bidirectional translation between 3D point clouds and 2D display modes (RGB, surface map, intensity map). Using cloud computing, online detection process runs extremely fast, around 45 seconds for a 40-million-pixel scan. Notification by email is sent when search is complete.

Note: there could be only one AI search process at a time on a given project. If one AI search has started already, notification about Tag list has pending AI jobs message will be shown.

Mind the warning message when working with Pipes/Cylinders. Certain restrictions are put on pipe diameters based on the algorithm and scan coverage.

5. Once search process is complete, go through the validation step. Each tag detected automatically comes with confidence metadata that can be found in Properties. Each tag created by AI Detection automatically comes with the status Validation needed.

6. Select Go to to view the selected tag in the closest scan location computed by Cintoo.

Add metadata or adjust the bounding box for the selected tag manually by clicking on Edit in the list of possible actions for a tag.

7. Click Validate to validate the tag or click Delete to remove the tag.

8. Repeat steps 5-7 for each tag in the list to be validated.