Turning accurate raw data into rich, usable insights – the Spatiall AI story
The exciting reality of AI has reached the shores of the geospatial data industry with a vengeance. AI-InfraSolutions has developed an end-to-end solution called Spatiall to provide users with valuable, actionable insights from accurate geospatial data. One of the key benefits of this end-to-end model is that the AI is trained on data that we have collected ourselves. Since we control that entire process, we can rely on the data and ensure its compatibility for optimum AI analysis. This is a USP.
AI pipeline to value
It all begins with a core dataset that is reliable and as complete as possible. AI algorithms need data that’s structured, validated, and available in compatible formats. So, we use our own Spatiall Data, panoramic images and LiDAR point clouds collected through our own mobile mapping and refinement processes. The data is prepared for optimal AI analysis – for example panoramic images are cut into virtual front, back, and side images.
AI scripts need to be trained, and this starts with geospatial experts labelling assets extracted from the images. They annotate what they are, how they appear from different angles, and relevant parameters for each asset type.
For example, for traffic signs, the AI model has been trained to recognize and identify: shape, type, dimensions, function, text, facing/pointing direction, carrier type, and carrier dimensions. These descriptors add a whole new dimension of information to add to the asset’s location.
Each asset type has a precise location attribute, However, since users need different information for different assets, each type has different description attributes. Machine Learning Engineers create process pipelines for each asset type.
Quality equals confidence
Of course, any data process needs to be quality checked. In this case, the raw data is already quality checked through our Spatiall Data refinement process. This checks the validity of the asset and the accuracy of its location, for example.
Then, our Engineers carry out manual reviews to validate AI outputs, especially for complex assets, edge cases (for example unusual assets or challenging environmental conditions), or high-value datasets. These reviews are also tailored to need – whether responding to customer requirement, user feedback, or asset type complexity.
Engineers use a list of standards and subjective performance metrics, including detection and identification accuracy, precision and recall, and confidence scores per asset.
Samples of AI-generated assets are also compared with ground truth data by geolocation experts. This means that asset types, locations, and meet defined quality thresholds.
All of these quality measures are designed to deliver the best data product to customers, giving them confidence to make the right decisions.
Informed decisions – in practice
Insights from Spatiall AI Data give users a wealth of asset information from the comfort of their desk. They can make decisions based on actual scenarios and make them efficiently – there’s no need for time-consuming and costly site visits. By highlighting critical assets or anomalies, AI helps users to prioritize resources and improve efficiency.
An example was voiced by in the Netherlands, who was able to manage his light poles portfolio efficiently using up-to-date information from AI-InfraSolutions. Check out his story here.
Our end-to-end model ensures that AI outputs are more accurate, scalable, and adaptable to real-world conditions, while giving customers a flexible solution that fits seamlessly into their existing systems.
The Spatiall AI Data can be added to our Spatiall Studio platform or delivered in compatible formats to integrate into existing customer systems.
Insights from Spatiall AI Data give users a wealth of asset information from the comfort of their desk. They can make decisions based on actual scenarios and make them efficiently – there’s no need for time-consuming and costly site visits. By highlighting critical assets or anomalies, AI helps users to prioritize resources and improve efficiency.
An example was voiced by in the Netherlands, who was able to manage his light poles portfolio efficiently using up-to-date information from AI-InfraSolutions. Check out his story here.
Our end-to-end model ensures that AI outputs are more accurate, scalable, and adaptable to real-world conditions, while giving customers a flexible solution that fits seamlessly into their existing systems.
The Spatiall AI Data can be added to our Spatiall Studio platform or delivered in compatible formats to integrate into existing customer systems.
Evolution of Spatiall AI
The value of AI lies in its constant improvement. The more data it processes, the better it gets at analyzing that data. And this means the accuracy and usefulness of the insights derived from it also improve.
And when data gets to high levels of accuracy, it becomes even more powerful. It will not only be able to detect and enrich assets but also provide predictive insights, flag anomalies before they become issues, and support real-time decision-making.
It’s already possible to see where there’s a gap in traffic signage by comparing datasets. But imagine, for example, being able to predict when you will need to replace or repair a light pole before it goes dark. Or which traffic signs and road markings are most likely to be found in a particular place, based on the layout of hundreds of similar locations.
Ultimately, AI makes data faster, more scalable, and increasingly intelligent. So its central to helping users turn complex mapping data into actionable insights more efficiently than ever. And its all part of an integrated end-to-end spatial data process.
Find out more about Spatiall AI