Data Sources

Lidar
Vehicle Imagery
GIS
Aerial Imagery
Drone Imagery
Cadaster

Data Model & Workflow

Deployment
Annotations
REST-API
Anonymization
Viewer
Model Training

Use Cases

Asset Management
Condition Monitoring

Objects & Surfaces

Traffic Signs
Roads
Road Markings

Detekt Platform Overview

Geospatial data collected from a mobile vehicle is the starting point for Detekt. Our platform empowers you to efficiently manage public space infrastructure, create and maintain digital inventories and automatically map any object and surface with Artificial Intelligence.

Smart city planning
Provide useful data for smart city planning, mobility and security
New insights into city infrastructure
Gain new insights into city infrastructure
City asset management
Support city asset management and optimize asset life-cycle costs
Train your own AI models
Train your own AI models to map objects which exist exclusively in your city or country
Object measurement in public space
Measure any detected object or surface with highest precision
Detekt AI Platform
Connect existing tools and software via API to further process detections
Next Section
Next Section

Data Sources

Lidar

Lidar (Light Detection and Ranging) is a remote sensing method that uses light in the form of a pulsed laser to measure ranges (variable distances) and to generate three-dimensional information about the shape of an area or object. We are using Lidar combined with GPS information to determine the exact location and size of a detected object.

Vehicle Imagery

Urban scenes are often captured by vehicles with a mobile mapping system using either a number of synchronized cameras with different orientations or one 360° camera. The result is always a set of georeferenced images which also allow precise measurements in case of stereoscopic images or an image/Lidar combination.

GIS

GIS is a standard of encoding geographical information into a file. We integrate existing GIS data into our models as annotations to support and improve model training and performance.

Aerial Imagery

Similar to vehicle imagery, high resolution aerial images taken by satellites or planes allows object detection and localization if combined with Lidar or if multi-view images are present.

Drone Imagery

Images from UAVs can compete with traditional aerial mapping solutions if they are post-processed in order to remove positional inaccuracies and to improve imagery orientation information.

Cadaster

Similar to GIS, cadaster data provides information on urban shapes and can therefore be used as input data for model training.
Next Section
Next Section

Data Model & Workflow

Next section

Deployment

Detekt can be implemented and run on premises and does not need internet access to function. This is especially relevant when working with sensible data. If needed, it can also be hosted on either AWS, Gcloud or Azure.

Annotations

Annotations in images or Lidar data are the basis for any object detection model. We can either integrate existing annotations or provide third-party support to create new object classes.

REST-API

The API allows integration into asset management applications to easily compare detected and stored objects.

Anonymization

Our partner Celantur provides a high-quality and scalable image anonymization solution to blur faces, bodies and license plates on panoramas, planar images and videos which we can fully integrate in our pipeline.

Viewer

The viewer allows a more visual representation of detected objects and integrates & simplifies the current annotation process.

Model Training

For continuous improvement, detections can be adapted and models retrained. In case annotations and models do not exist for your desired use case yet, annotations can be made by yourself to train a new model from scratch.
Next Section

Use Cases

Asset Management

Counting and managing all assets is key to make a city operate. With Detekt, you can leverage your available data for that matter. Any visible object can be detected, measured and its exact position determined to match this information with existing asset management databases.

Condition Monitoring

An asset which is damaged or shut down for maintenance is often difficult to identify and costly. Traditional on-site asset inspection is very time-consuming and inefficient, even if mobile mapping image data is available to perform this task from your desktop. Detekt analyzes existing footage using Artificial Intelligence to identify and classify defects, be it a crack or a damaged traffic sign. Maintenance staff no longer need to manually look for defects or rely on reports from observant citizens and benefit from time savings and an improved maintenance process.

Objects & Surfaces

Traffic Signs

Traffic sign detection and localization currently works for any sign worldwide with classifications available for Austria. With little annotation effort provided by you or our annotators, traffic sign classification can be performed within any city or country worldwide.

Roads

The following elements can be detected and classified on public roads, bike paths and sidewalks:

  • Various defects such as cracks, alligator cracks, open joints, pot holes and substantial damages
  • Surfaces such as paving stones, concrete pads, rails, track beds, manhole covers
  • Temporary and permanent patches
  • Repaired cracks and closed joints

Road Markings

Markings drawn on the road (such as arrows, turn only lane signs, speed limits, etc.) can be identified for a complete overview.