Offer a solution for a pilot call

Mobility Lab Helsinki is looking for solutions for the development of a digital twin for mobility.

We are looking for solutions for agile pilots that explore and test new methods, such as artificial intelligence, for the information modelling of the current state of street environments and street asset management. The aim is to serve the development of the city’s digital twin.

Fill out the form here (in English)

Fill out the form here (in Finnish)

The call is open until 16:00 on 4 December 2023.

Forum Virium Helsinki will carry out the pilot procurement as part of the Mobility Lab Helsinki project.

The new solutions to be tested can relate to themes such as the following:

1. a method for identifying and classifying street traffic conditions or equipment as 3D geospatial data, for example by using new AI methods to analyse the point clouds of a mobile laser scan to identify lanes and street equipment
2. a method for generating line data on the centre lines of traffic lanes, for example by developing an identification method based on point clouds from mobile laser scanning for an open source application
3. a method for supporting street equipment maintenance processes, such as a crowdsourcing traffic sign validation using mobile application
4. a method for combining existing geospatial data into useful new data, such as using mobile laser scans of the street space to refine street spaces in the city model
5. development of the automatic classification of the point clouds of street & airborne laser scanning, for example by crosschecking the consistency and quality of the classification of different datasets
6. improving the combined viewing of street view and city model data, for example by developing methods for presenting 360 images on city model presentation platforms (Cesium).

Take a look at the invitation to tender in Finnish

Take a look at the invitation to tender in English

The open invitation to tender is seeking a maximum of three suppliers capable of meeting the needs described above.

The total budget of the procurement is up to EUR 45,000. The maximum cost (price ceiling) of a single pilot is EUR 15,000, VAT 0%.

The call for tenders will be open until 16:00 on 4 December 2023.

A Q&A section with freguently asked questions will be released on this page at a later date.

Q&A

What is the nature of open source in pilot projects?

The aim is certainly to produce open source code, but it’s acknowledged that this might not always be feasible in every situation. The crucial aspect is to clearly state in your proposal which components will be released as open source results and which will contain trade secrets and proprietary business logic that cannot be shared. The nature of the pilot call is to share as many results as possible, with the hope of benefiting the entire field.

Has the issue of recognising traffic signs from street spaces been resolved, or is this still a relevant topic in this pilot call?

In last autumn’s pilot, traffic signs were analysed from mobile laser scan & 360-photo data. The city has also purchased traffic signs recognition as a part of their official street data collection campaign during summer 2023.. These signs will be accessible via the city’s open WFS APIby the end of the year. If you have innovative approaches in how to validate traffic sign detections from different datasets, you should include them in your proposal.

Latest information regarding traffic sing detections in Helsinki is well described in Pekka Koponen’s presentation https://youtu.be/PCJ8L7FqsNY?si=lYvU4FROEwt2pYIh&t=1748

Is CityInfra operating on the Cesium platform? 

No, currently Cityinfra does not have any dedicated public viewing service. Cityinfra is providing data out through City’s open WFS API (layers: additional_sign_* and traffic_sign_*) and their own REST-API (https://city-infra.test.hel.ninja/swagger/. However, it is acknowledged that Cesium is the common nominator in bringing together different kinds of 3D datasets for visualisation. As we are working with this 3D mobile laser scanning dataset, we aim for the final presentation to be in 3D format, should we proceed in this direction.

Is street view visualisation straightforward, or are there additional user stories that need to be considered? 

A key point here is that the city has a lot of different users and use cases for the acquired data. Cyclomedia has provided their StreetSmart platform for viewing and exporting the city wide dataset. City has several different GIS-based services, where a viewing component could be integrated. Having an open source alternative could come in handy also for more specialised workflows. As mentioned before, such open source components could be based on Cesium, as it’s a common nominator on which multiple different 3D-web platforms are based. Viewing street view data in QGIS is one method, however Cyclomedia has an official plugin for QGIS https://www.cyclomedia.com/en/qgis.

Is street view imaging the primary focus for user visibility, or do point clouds, 3D tiles, and other 3D dimensions play a significant role?

This question cannot be fully answered at this moment. 3D tiles, a format for feeding data to Cesium, are widely used. While images and point clouds both have their applications, images may require more careful consideration from a privacy standpoint. Therefore, point cloud data is generally more straightforward to use across various systems.

Regarding training models, is there already labelled data available for an AI algorithm to recognise elements in 360-degree images?

No, there are not. From our analysis of the dataset, it may not be suitable for training purposes, such as identifying potholes in streets, because it contains a very limited number of examples. We recommend utilising open foundational AI models, like those released by Meta in the past year. Implementing solutions based  on pre-trained models could offer novel insights, perhaps even by benchmarking their effectiveness when used in Helsinki’s street view context. It’s possible that these models may already recognise elements within the data, which would eliminate the need for training new models from the scratch.

The provided mobile laserscan dataset, is the point cloud already classified? Is it available in different formats? 

Source dataset is not yet classified, data is provided as is we got it from Cyclomedia. In addition to original LAZ-files, the point cloud is available as 3DTiles point cloud. This data can be previewed at https://proto.fvh.io/jatkasaari/HKI-vs-cyclomedia.html and can be provided to a companies selected for the pilot either as files or accessable directly from Forum Viriums Azure blob.

Helsinki city model, which kind of versions and datasets are available of it?

There are several types of datasets available as 3DTiles, mostly buildings. City information model: textured and untextured LOD2 buildings, ground terrain (Cesium terrain). Photogrammetry mesh: Years 2015 and 2017. 2017 is also available as 3DTiles glTF 2.0 version, which is supported by certain engines running Cesium. Basically, everything you can see in the https://kartta.hel.fi/3d service side menu, are available as 3DTiles. More information https://hel.fi/3d or https://www.hel.fi/en/decision-making/information-on-helsinki/maps-and-geospatial-data/helsinki-3d

We are getting tenders from Helsinki KYMP’s “Paikkatietoasiantuntija DPS” dynamic procurement system, and noticed that in one of recent tenders there seems to be overlap with the goals of this open call. Can you elaborate how these tenders are aligned?

Yes, there are partly overlapping tenders currently active. We are actively coordinating with KYMP not to acquire overlapping work. Please don’t send the same offer for our open call as to KYMP open tenders.

Information event on 27 November at 13:00

Download information webinar presentations (pdf)

Recording of the information webinar on November 27 2023.