Feed on

Since 2010 organized humanitarian mapping has evolved as a constant and growing element of the global OpenStreetMap (OSM) community. With more than 7,000 projects in 150 countries humanitarian mapping has become a global community effort. Mappers have added more than 60 Million buildings to OSM through HOT’s Tasking Manager. That’s around 13% of all OSM building contributions since 2008! However, validation still is a challenge for humanitarian mapping activities in OSM. Validation could catch up with mapping in only 3 month (out of 91) since 2012.

At HeiGIT we have been developing the new Humanitarian OSM Stats tool and dashboard to analyze and visualize stats and facts on the global scale and spatial footprint of (humanitarian) mapping in OSM such as the ones above. You are invited to take a look at the dashboard and explore statistics for mapping projects in HOT’s Tasking Manager. A key component of our analysis lies in the evaluation of the data contributed to OSM. A contributions is defined as any change in an OSM element and can be assessed based on OSM full-history data. We use the ohsome OpenStreetMap History Analytics platform, a high-performance spatio-temporal data analysis platform for OpenStreetMap full-history data developed by HeiGIT for this analysis based on top of HeiGIT’s OSM history database OSHDB.

We are planning to open-source the code used for the analysis soon. We are in close contact with members of HOT and the further OSM community and are constantly exploring, how we can add further aspects to the analysis. Feel free to reach out to us, e.g. via email info@heigit.org, in case you have specific questions or ideas on what would make the dashboard even more useful.

Related Work

Herfort, Benjamin; Raifer, Martin; Reinmuth, Marcel; Stier, Jochen; Klerings, Alina (2020): Evolution of humanitarian mapping within the OpenStreetMap Community. Talk at the State of the Map conference 2020, Cape Town.

Raifer, Martin; Troilo, Rafael; Kowatsch, Fabian; Auer, Michael; Loos, Lukas; Marx, Sabrina; Przybill, Katharina; Fendrich, Sascha; Mocnik, Franz-Benjamin; Zipf, Alexander (2019): OSHDB: a framework for spatio-temporal analysis of OpenStreetMap history data. Open Geospatial Data, Software and Standards.

Auer, M.; Eckle, M.; Fendrich, S.; Griesbaum, L.; Kowatsch, F.; Marx, S.; Raifer, M.; Schott, M.; Troilo, R.; Zipf, A. (2018): Towards Using the Potential of OpenStreetMap History for Disaster Activation Monitoring. ISCRAM 2018. Rochester. NY. US.

Auer, Michael; Eckle, Melanie; Fendrich, Sascha; Kowatsch, Fabian; Marx, Sabrina; Raifer, Martin; Schott, Moritz; Troilo, Rafael; Zipf, Alexander (2018): Comprehensive OpenStreetMap History Data Analyses- for and with the OSM community. Talk at the State of the Map conference 2018, Milan.

Raifer, Martin (2017): OSM history analysis using big data technology. Talk at the State of the Map conference 2017, Aizuwakamatsu.

Scholz, S., Knight, P., Eckle, M., Marx, S., Zipf, A. (2018): Volunteered Geographic Information for Disaster Risk Reduction: The Missing Maps Approach and Its Potential within the Red Cross and Red Crescent Movement. Remote Sens., 10(8), 1239, doi:10.3390/rs10081239.

A new month - a new post of the how to become #ohsome blog series. Welcome to part 7, in which we will show you 7 different ways how you can send a request to the ohsome API using different tools and programming languages. In case this is your first blog of how to become ohsome (and awesome), check out the collection of all posts of this series.

As you probably know, the ohsome API allows non-programmers to analyze the rich data source of the OpenStreetMap (OSM) history. There are a lot of ways to perform analysis using the ohsome API. For the purpose of this article, let’s assume we want to calculate the perimeter of residential buildings in a small area of Heidelberg.


The first two ways of accessing the ohsome API are via using GET and POST requests through cURL. It is a widely known command line tool for getting and sending data through URL addresses. Open the command line interpreter of your operating system and digit the following command:

curl -X GET "https://api.ohsome.org/v1/elements/perimeter?bboxes=8.681609,49.414517,8.690232,49.417254&filter=building=residential&format=json&showMetadata=true&time=2014-01-01/2017-01-01/P1Y&timeout=30" -H "accept: application/json"

For a POST request, just use the following command:

curl -d @query -X POST https://api.ohsome.org/v1/elements/perimeter

query is a text file, which contains the query string:


Another blog post, where we’ve used cURL to perform our analysis can be found here.


For the programmers among you, we have two more GET and POST examples using python. The GET request will look here as follows:

import requests
URL = 'https://api.ohsome.org/v1/elements/perimeter'
params = {"bboxes": "8.681609,49.414517,8.690232,49.417254", "filter": "building=residential", "format": "json", "showMetadata": "true", "time": "2014-01-01/2017-01-01/P1Y", "timeout": "30"}
response = requests.get(URL, params=params)

And analogous a POST request:

import requests
URL = 'https://api.ohsome.org/v1/elements/perimeter'
data = {"bboxes": "8.681609,49.414517,8.690232,49.417254", "filter": "building=residential", "format": "json", "showMetadata": "true", "time": "2014-01-01/2017-01-01/P1Y", "timeout": "30"}
response = requests.post(URL, data=data)

You like Python? Then you should definitely check out this post, where we’ve used a Jupyter Notebook to query the API and directly visualize the returned results.


Powerful statistical analysis and graphics: those are two aspects that one could associate with R. A GET request to the ohsome API in this programming language would look as follows:

r <- GET("https://api.ohsome.org/v1/elements/perimeter?bboxes=8.681609,49.414517,8.690232,49.417254&filter=building=residential&format=json&showMetadata=true&time=2014-01-01/2017-01-01/P1Y&timeout=30")

The code for a POST request:

r <- POST("https://api.ohsome.org/v1/elements/perimeter", encode = "form", body = list(bboxes = "8.681609,49.414517,8.690232,49.417254", filter = "building=residential", format = "json", showMetadata = "true", time = "2014-01-01/2017-01-01/P1Y", timeout = "30"))


The 7th and last way that we want to explain you here how to access the ohsome API is through our swagger UI. It’s very suitable if you want to get an overview of all the available endpoints, or just make a query in a simple way.

First of all, select a spec that defines whether you want a data aggregation, a data extraction or just the metadata. In our case we stick with the pre-selected Data Aggregation spec.

Next, you choose which kind of resource you want. Since we want to make a GET request for the resource /elements/perimeter, we select Perimeter, then GET /elements/perimeter After that, just click on Try it out.

Insert the values in the text area of the parameters that you want to use and click on “Execute”. You should get the requested data inside the response body window. Further, you can download the response through the download button on the bottom right.

Now you’ve seen seven different ways of how to speak to the ohsome API: GET and POST using cURL, python and R, as well as through the Swagger UI. This post only showed you one endpoint of many. Please head to our fresh documentation page (which got introduced here by the way) if you’re interested in knowing more about what the ohsome API has to offer. You have some further questions - contact us via info(at)heigit.org.

Background info: the aim of the ohsome OpenStreetMap History Data Analytics Platform is to make OpenStreetMap’s full-history data more easily accessible for various kinds of OSM data analytics tasks, such as data quality analysis, on a global scale. The ohsome API is one of its components, providing free and easy access to some of the functionalities of the ohsome platform via HTTP requests. Some intro can be found here:

So it seems, we need to repeat an updated version of a Blogpost from exactly one year ago:

You are looking for a more shady pedestrian route through the urban jungle? You might then prefer some routes that go through public green spaces with shady trees and bushes. Thank goodness we are working on such green and also shady routing together with several partners in the mFund project meinGrünGIScienceHD/HeiGIT have already some multi-criteria routing prototypes related to e.g. green routing preferring vegetated areas using OpenStreetMap data and other sources, but also on routes avoiding noise and preferring social activities.

Also the dedicated shady route” feature will soon be released for Heidelberg and Dresden in the meinGrün App as an additional routing criteria based on openrouteservice. It considers shade provided by trees and also high buildings.

Hope you enjoy the detours. Stay cool!

Novack, T.; Wang, Z.; Zipf, A. (2018): A System for Generating Customized Pleasant Pedestrian Routes Based on OpenStreetMap Data. Sensors 2018, 18, 3794.


Ludwig, Christina ; Zipf, Alexander (2019): Exploring regional differences in the representation of urban green spaces in OpenStreetMap. Proceedings of the GeoCultGIS - Geographic and Cultural Aspects of Geo-Information: Issues and Solutions, Limassol (Cyprus)

H. Tost, M. Reichert, U. Braun, I. Reinhard, R. Peters, S. Lautenbach, A. Hoell, E. Schwarz, U. Ebner-Priemer, A. Zipf, and A. Meyer-Lindenberg (2019): Neural correlates of individual differences in affective benefits of real-life urban green space exposure. Nature Neuroscience. https://doi.org/10.1038/s41593-019-0451-y

Z. Wang, T. Novack, Y. Yan and A. Zipf, “Quiet Route Planning for Pedestrians in Traffic Noise Polluted Environments,” in IEEE Transactions on Intelligent Transportation Systems, doi: 10.1109/TITS.2020.3004660.

Can the human brain be described as a kind of machine – and, by extension, human memory as a time machine? What do software and the human mind have in common, and what are the ethical dilemmas involved in the use of artificial intelligence (AI)? Neurobiologist Hannah Monyer and geoinformatics expert Bernhard Höfle present talking points on the relationship between man and machine, reflect on what tasks each can perform better and explain why algorithms alone do not produce perfect results.

The human brain – particularly its memory function – has always been associated with technology, says Hannah Monyer. But according to her, this comparison is flawed because unlike audio tapes or computers, the brain does not save information in precisely the way it occurred. Instead the brain sorts information during the learning process, saving some things and deleting others. “Our brain is not just a storage device for facts; the facts are selected even while they are registered and then undergo a process that is shaped by all that came before – every piece of information is put in a specific context.” She explains that neurobiologists owe much of their knowledge about the brain to advances in information technology – in many cases, it was computers that allowed scientists to learn how the brain works.

The field of geoinformatics combines the strengths of man and machine to answer questions of a geographical nature. According to Bernhard Höfle, machines can repeat clearly defined tasks such as measurements with the aim of achieving increased objectivity, while algorithms allow scientists to apply specific tasks to large digital data sets that would be impossible to process for humans. “On the other hand, humans are clearly in the lead when it comes to discovering things that we were not even looking for or could not conceive of.” He emphasises that with respect to AI, we need to define which tasks should be delegated to machines – with due regard to ethical and moral considerations – instead of all the things that could be delegated: “Not everything that is theoretically possible should ultimately be put into practice.”

Interview with Prof. Monyer (Neurobiology) and Prof. Höfle (Geography): https://heiup.uni-heidelberg.de/journals/index.php/rupertocarola/article/view/24184/17900

Read the full magazine online (in German): https://www.uni-heidelberg.de/de/presse-medien/publikationen/forschungsmagazin/maschine-mensch

Since some time the City of Karlsruhe uses the API of Openrouteservice (ORS) by HeiGIT for routing of pedestrians, bicycles and cars in the new online city plan and the citizen GIS app for the general public, i.e. the citizens and visitors of Karlsruhe. You can use it here:


The specific client developed by the Liegenschaftsamt, city of Karlsruhe is based on the ArcGIS for Javascript API and optimises as a widget for the ArcGIS Web AppBuilder by Esri. The public transportation functionality is based on the TRIAS interface.

This demonstrates that Openrouteservice can easily be integrated in several clients and apps. Openrouteservice uses the free and open data of OpenStreetMap (OSM) and provides a rich set of API functionality for the whole globe, as well as an ecosystem of different libraries and clients, e.g. for Javascript, Python, R, QGIS, Java etc.

The list of the free and open source ORS API features include, e.g. also

  • routing with directions for pedestrian, wheelchair, car, different types of bicycle profiles and heavy vehicle profiles with many options
  • vehicle fleed optimization (traveling salesman etc)
  • isochrones for reachability applications
  • time-distance matrices
  • geocoding and reverse geocoding
  • points of interest

Please note that the quota for openrouteservice multi vehicle optimization has been increased to support logistics during the Corona crisis. You can also test the interactive documentation of the different API endpoints. Recent research at GIScience Heidelberg and HeiGIT related to routing deals e.g. with healthy and green routing, or quite routing, as well as applications in humanitarian aid (e.g. access to healthcare in Africa).

Physical activity is beneficial for human physical health and well‐being. Accordingly, the association between physical activity and mood in everyday life has been a subject of several Ambulatory Assessment studies. This mechanism has been studied in children, adults and the elderly, but neglected in adolescents. It is critical to examine this mechanism in adolescents because adolescence plays a key role in human development and adolescents’ physical activity behavior translates into their behavior in adulthood.

In a recently accepted study we investigated adolescents’ mood in relation to distinct physical activities: incidental activity such as climbing stairs; exercise activity, such as skating; and sports, such as playing soccer. We equipped 134 adolescents aged 12‐17 years with accelerometers and GPS‐triggered electronic diaries to use in their everyday life. Adolescents reported on mood repeatedly in real‐time across seven days and this data was analyzed using multilevel‐modeling.

Results in short are: After incidental activity, adolescents felt better and more energized. After exercise, adolescents felt better but less calm. After sports, adolescents felt less energized. Analyses of the time course of the effects confirmed our findings.

Physical activity influences mood in adolescents’ everyday life, but has distinct effects depending on the kind of physical activity. Our results suggest incidental and exercise activities entails higher post‐bout valence compared to sports in competitive settings. These findings may serve as an important empirical basis for the targeted application of distinct physical activities to foster well‐being in adolescence.

Koch E. D., H.  Tost, U. Braun, G. Gan, M.  Giurgiu, I. Reinhard, A.  Zipf, A.  Meyer‐Lindenberg, U.  Ebner‐Priemer, M.  Reichert (2020): Relationships between Incidental Physical Activity, Exercise, and Sports with subsequent Mood in Adolescents. The Scandinavian Journal of Medicine & Science in Sports. https://doi.org/10.1111/sms.13774

Related work:

Reichert, M., Braun, U., Lautenbach, S., Zipf, A., Ebner-Priemer, U., Tost, H., Meyer-Lindenberg, A. (2020): Studying the impact of built environments on human mental health in everyday life: methodological developments, state-of-the-art and technological frontiers. Current Opinion in Psychology 32, 158-164.

Tost, H., Reichert, M., Braun, U., Reinhard, I., Peters, R. , Lautenbach, S., Andreas, H., Schwarz, E., Ebner-Priemer, U., Zipf, A., Meyer-Lindenberg, A. (2019): Neural correlates of individual differences in affective benefit of real-life urban green space exposure. Nature Neuroscience. https://doi.org/10.1038/s41593-019-0451-y

Novack, T., Wang, Z., Zipf, A. (2018): A System for Generating Customized Pleasant Pedestrian Routes Based on OpenStreetMap Data. Sensors 2018, 18, 3794. https://doi.org/10.3390/s18113794

Reichert, M., Giurgiu, M., Koch, E., Wieland, L.M., Lautenbach, S., Neubauer, A.B., von Haaren-Mack, B., Schilling, R., Timm, I., Notthoff, N., Marzi, I., Hill, H., Brüßler, S., Eckert, T., Fiedler, J., Burchartz, A., Anedda, B., Wunsch, K., Gerber, M., Jekauc, D., Woll, A., Dunton, G.F., Kanning, M., Nigg, C.R., Ebner-Priemer, U., Liao, Y. (2020): Ambulatory assessment for physical activity research: State of the science, best practices and future directions. Psychology of Sport and Exercise 101742. https://doi.org/10.1016/j.psychsport.2020.101742

Törnros, T., Dorn, H., Reichert, M., Ebner-Priemer, U., Salize, H.-J., Tost, H., Meyer-Lindenberg, A., Zipf, A. (2016): A comparison of temporal and location-based sampling strategies for GPS-triggered electronic diaries.” Geospatial Health. Vol 11, No 3. DOI:10.4081/gh.2016.473.

Reichert, M., Törnros, T., Hoell, A., Dorn, H., Tost, H., Salize, H.-J., Meyer-Lindenberg, A., Zipf, A., Ebner-Priemer, U. W. (2016). Using Ambulatory Assessment for experience sampling and the mapping of environmental risk factors in everyday life. Die Psychiatrie. 2/2016. 94-102. (pdf)

End of last week Dr. Xuke Hu successfully defended his PhD thesis entitled “Building Semantics Reasoning by Using Rules based on Available Geospatial Information“.

His dissertation investigated the potential of inferring distinct indoor and outdoor spatial (specifically building) elements based on existing or available spatial elements on OSM or provided by sensing equipment, leveraging the association relationship between the spatial elements. Furthermore, this dissertation compared and explored the applicability of two kinds of reasoning mechanisms using manually defined explicit rules (Knowledge driven) and learned implicit rules (data driven) in this context, respectively. Four representative indoor and outdoor building elements (i.e., roof shape, room usage, main entrance, and landmark salience) were taken as examples to explore how and why the four building elements can be inferred by explicit and/or implicit rules. Finally, the results of the four studies were combined to answer the questions related to the research objectives.

The PhD consists of a introduction and synopsis and the following five journal paper:

  1. Hu, X., Fan, H., Noskov, A. (2018): Roof model recommendation for complex buildings based on combination rules and symmetry features in footprints. International Journal of Digital Earth. Vol. 11(10), pp.1039-1063, Doi: 10.1080/17538947.2017.1373867.
  2. Hu, X., Noskov, A., Fan, H., Novack, T., Gu, F., Li, H., Shang, J.: Tagging the Buildings’ Main Entrance based on OpenStreetMap and Binary Imbalanced Learning. (revision)
  3. Hu, X., Fan, H., Noskov, A., Zipf, A., Wang, Z., Shang, J. (2019). Feasibility of Using Grammars to Infer Room Semantics. Remote Sensing. vol. 11(13), p.1535. Doi: 10.3390/rs11131535.
  4. Hu, X., Fan, H., Noskov, A., Wang, Z., Zipf, A., Gu, F., Shang, J. (2020 accepted). Room Semantics Inference Using Random Forest and Relational Graph Convolutional Network: A Case Study of Research Building. Transactions in GIS. (accepted)
  5. Hu, X., Ding, L., Shang, J., Fan, H., Novack, T., Noskov, A., Zipf, A. (2019). A Data- driven Approach to Learning Saliency Model of Indoor Landmarks by Using Genetic Programming. International Journal of Digital Earth. https://doi.org/10.1080/17538947.2019.1701109

Because of the corona pandemic the external reviewer Prof. Dr. WenWen Li from Arizona State University unfortunately could only participate virtually. Originally she had planned to stay her sabbatical this semester with us at the GIScience Research Group Heidelberg University.

We congratulate Xuke most cordially for this achievement! We are looking forward to further joint work. All the best at your new job at DLR Data Science Institute, Jena!

As you may already know from the latest blog posts, new advancements are brought with the release 1.0 of the #ohsome API. One minor function involves the correctness of parameters that you can use in your queries, which was developed by our intern Rosario in the last weeks.

The new feature

Let’s say you send a request to compute the amount of residential buildings in a small area in Heidelberg on 1st of January 2015. You misspell the parameter “bboxes” and write “bbox” instead. In this case, the new feature of the ohsome API will give you a quick response back, indicating that the parameter “bbox” is unknown and it will suggest – if possible – the proper parameter to use. The response of the ohsome API will look as follows:

"timestamp": "2020-07-13T14:58:05.346",
"status": 400,
"message": "Unknown parameter 'bbox' for this resource. Did you mean 'bboxes'?",

You could also use a parameter which is correct, but is supposed to be used for other resources. Let’s suppose you make this request. Although the ohsome API knows about the parameter “properties”, its use in this context is wrong, since “properties” doesn’t work for /elements/count (here you can find a list about which parameter works for which resource). The ohsome API will give the following message back:

“message”: “Unknown parameter ‘properties’ for this resource.”

So don’t be afraid to be wrong, the ohsome API will help you!

How it works

The new implemented feature uses a string similarity algorithm called “Fuzzy Score”: When an unfit parameter is encountered, it is compared against the list of correct parameters for the requested resource of the ohsome API. If a sequence of characters matches against a sequence of characters of a proper parameter in the list, then the algorithm will give points to this parameter. A higher final score indicates that more characters are in sequence, which implicates more similarity. Only the first two parameters with the highest score will be suggested in the response. Each of them must have at least a score of five. If the score is lower than five, you will get only an “unknown parameter” message without suggestion. We chose five because according to the behavior of the algorithm and to our set of valid parameters, a score of five is suitable as threshold to either give a suggestion (score >= 5) or don’t (score < 5). Furthermore, the algorithm is case insensitive, which means it works regardless if the parameter is written in upper or lower case letters.

Before writing the code we valuated a lot of string similarity algorithms which could have helped for our purpose. If you want to know more about it, you can start by checking this easy introduction. Also if you want to know how this feature was implemented, check the code on GitHub: StringSimilarity.java and ResourceParameters.java.

ohsome sounds like awesome

Background Info

The aim of the ohsome OpenStreetMap History Data Analytics Platform is to make OpenStreetMap’s full-history data more easily accessible for various kinds of OSM data analytics tasks, such as data quality analysis, on a global scale. The ohsome API is one of its components, providing free and easy access to some of the functionalities of the ohsome platform via HTTP requests. Some intro can be found here:

GIScience methods are becoming more and more widespread in different domains. A current review article under involvement of HeiGIT researchers discusses the potential of ambulatory assessments and GIScience approaches for physical activity research. The article is on of the outcomes of the 2nd International CAPA Workshop 2019Physical Activity Assessment – State of the Science, Best Practices, Future Directions” at KIT.

The availability and widespread use of wearable devices including pedometers, accelerometers and GPS devices offers a big potential for research, including observational, experimental and international studies. Research questions are not limited to the question of where and when physical activities are undertaking but involve the identification of motivating and demotivating factors as well as the question on the effects of mental and physical well-being. For such question longitudinal studies involving ambulatory assessment methods -  for example the use of e-diaries - are well suited.

The use of GIScience methods is not limited to establishing context by means of spatial intersections or spatial joins but could involve the use of specialized routing services for intervention studies. Observational studies have mainly used information on stressors or distresses along the path as explanatory factors. Routing services could be used for example to suggest paths to nearby urban green spaces for interventional exercises or could suggest routes for physical activities that vary in comfort (e.g. by degree of noisiness or greeness) to test effects on motivation and well being.

Reichert, M., Giurgiu, M., Koch, E., Wieland, L.M., Lautenbach, S., Neubauer, A.B., von Haaren-Mack, B., Schilling, R., Timm, I., Notthoff, N., Marzi, I., Hill, H., Brüßler, S., Eckert, T., Fiedler, J., Burchartz, A., Anedda, B., Wunsch, K., Gerber, M., Jekauc, D., Woll, A., Dunton, G.F., Kanning, M., Nigg, C.R., Ebner-Priemer, U., Liao, Y. (2020): Ambulatory assessment for physical activity research: State of the science, best practices and future directions. Psychology of Sport and Exercise 101742. https://doi.org/10.1016/j.psychsport.2020.101742
For the next weeks the article can be accessed freely using the following share link
The findings of the review article are closely linked to our activities in the psychogeography project and relate to the meinGrün project and proposed follow-up activities.
Related publications:

After more than a decade of rapid development of volunteered geographic information (VGI), VGI has already become one of the most important research topics in the GIScience community. Almost in the meantime, we have witnessed the ever-fast growth of geospatial deep learning technologies to develop smart GIServices and to address remote sensing tasks, for instance, land use/land cover classification, object detection, and change detection. Nevertheless, the lack of abundant training samples as well as accurate semantic information has been long identified as a modeling bottleneck of such data-hungry deep learning applications. Correspondingly, OpenStreetMap (OSM) shows great potential in tackling this bottleneck challenge by providing massive and freely accessible geospatial training samples. More importantly, OSM has exclusive access to its full historical data, which could be further analyzed and employed to provide intrinsic data quality measurements of the training samples.

Recently, we open-sourced our ohsome2label tool, which offers a flexible framework for labeling customized geospatial objects using historical OSM data that allows more effective and efficient deep learning. Based on the ohsome API developed by HeiGIT gGmbH, ohsome2label aims to mitigate the lack of abundant high-quality training samples in geospatial deep learning by automatically extracting customized OSM historical features, and providing intrinsic OSM data quality measurements.

Package description flowchart of ohsome2label.

The aim of this tool is to promote the application of geospatial deep learning by generating and assessing OSM training samples of user-specified objects, which not only allows user to train geospatial detection models but also introduces the intrinsic quality assessment into the “black box” of the training of deep learning models. With such deeper understanding of training samples quality, future efforts are needed towards more understandable and geographical-aware deep learning models.

Follow our further activities in GitHub, and stay tuned for more ohsome news: https://github.com/GIScience/ohsome2label

You are also invited to read our conference paper in proceeding of the Academic Track at the State of the Map 2020:

Wu, Zhaoyan, Li, Hao, & Zipf, Alexander. (2020). From Historical OpenStreetMap data to customized training samples for geospatial machine learning. In proceedings of the Academic Track at the State of the Map 2020 Online Conference, July 4-5 2020. DOI: http://doi.org/10.5281/zenodo.3923040

Find more information about ohsome API:

The ohsome OpenStreetMap History Data Analytics Platform makes OpenStreetMap’s full-history data more easily accessible for various kinds of OSM data analytics tasks, such as data quality analysis, on a global scale. The ohsome API is one of its components, providing free and easy access to some of the functionalities of the ohsome platform via HTTP requests. More information can be found here:

Earlier work and examples of deep learning and machine learning with training labels from OSM (and MapSwipe) includes e.g.:

Older Posts »