Said Turksever

Powerful earthquakes hit southern Turkey and Syria on 6 February 2023. These earthquakes in Turkey and Syria caused thousands of casualties and destroyed cities. Geospatial infrastructure is critical to respond to these earthquakes during rescue operations, humanitarian effort as well as planning recovery activities. Yercizenler coordinated mapping activation with the collaboration of Humanitarian OpenStreeMap team to improve open geodata infrastructure in the earthquake affected region and supporting humanitarian response in the scope of mapping. Türkiye Earthquakes Mapping Response aims to complete open map data infrastructure before and after the event in affected areas. This response is structured with following workstreams; Remote Mapping, Post-disaster Field Data Collection, Global Community Activation and Geo-data Integration. In this talk; we will talk about how open data and community activation helped save lives after earthquakes, what challenges we faced and what we have learnt during the Türkiye Earthquakes mapping Response effort.

None

Traffic signs are a key feature for navigating and managing traffic safely, affecting all of us on a daily basis. However, traffic sign datasets are lacking on open government data portals as well as OpenStreetMap (OSM). Mapillary’s computer vision capabilities can extract more than 1,500 classes of traffic signs globally from street-level imagery. Generated traffic signs are available on iD Editor, Rapid and JOSM Mapillary plugin to enrich OpenStreetMap data. Our team wanted to know how the accuracy of traffic signs detected by Mapillary compared with the reality on the ground (the ground truth). To answer this question we collected more than thousands ground truth data in San Francisco and used this information to produce the recall, precision, and positional accuracy of our machined generated traffic sign data. This provided some interesting insights in OpenStreetMap and the level of completeness and gaps of that dataset. In this talk, we will cover Mapillary’s traffic sign extraction capabilities, Mapillary generated traffic sign data against ground truth data and OSM’s traffic sign coverage in San Francisco’s downtown. We will be also addressing how data quality can be improved using various data collection techniques and the role of post-processing with Structure from Motion and control points annotations.

None

Other people: Yunzhi Lin

Mapping is time-consuming and requires a high workforce when it comes to keeping maps up-to-date. Mapillary brings a different approach to geospatial data collection with street-level imagery. This approach allows communities to collect geospatial data faster and cheaper. But can Mapillary-generated data be useful for enriching OpenStreetMap? In this study, Mapillary-extracted map data will be examined against ground truth to assess data quality on contributing to OpenStreetMap.

In this case study the positional accuracy and completeness are assessed by benchmarking Mapillary-generated data against ground truth for street light provided by Ordu Metropolitan Municipality in the borough of Altinordu, which covers an area of 9 km2. A total number of 323 street lights were recorded in the authoritative dataset. Completeness and positional accuracies are evaluated for each street light for different camera setups. In order to compare the impact of camera type on positional accuracy, street-level imagery was collected with three different cameras: iPhone 11, GoPro Hero 7 Black and GoPro Max. Collected street-level imagery was uploaded to Mapillary by using the Desktop Uploader. Imagery which are captured from different cameras is isolated during the upload to compare map data accuracy, completeness and correctness comparison based on camera type. In this experiment, we validate the effectiveness of Mapillary extracted map data by focusing on streetlights by evaluation results based on completeness and positional accuracy key performance indicators. The best result of completeness is achieved with GoPro Hero 7 Black with 87.57% in the working area and it is followed by GoPro MAX with 77.30% for Mapillary extracted street light data . Lastly, the completeness of iPhone 11 acquired data is 71.89%. In terms of positional accuracy, our experiment shows GoPro MAX captured street-level imagery can be extracted with 2.02 m of positional error and it is followed by GoPro Hero 7 Black with 2.17 m of positional error. The average positional error of street light which is extracted from iPhone 11 captured street-level imagery is 2.21 m. This positional error is close to the precision of a single frequency GPS receiver. This experimental study shows that positioning accuracy is highly related to the GPS accuracy of the capture device, and in general, a large part of the final positional error can be attributed to this. Mapillary's 3D reconstruction is able to mitigate some of these effects. Additionally, capturing with a large field of view has a positive effect on accuracy. In this study we also validate that positional accuracy depends on various factors of the capture process; precision of GPS receiver and additionally positioning hardware, resolution and quality of the images, image capture frequency of the camera, imagery density in the working area and type of photo such as flat or 360. The overall positional accuracies are under 5m which can be a promising solution for enriching street lights data on OpenStreetMap and collecting streetlights inventory for the municipalities and governmental bodies if this data will not be used for surveying purposes or reference data, however Mapillary-generated data can be useful and time effective as complementary data with low cost collection expenses.

Turkish OpenStreetMap Community has constantly been growing with members from various circles in recent years however the community faces various problems such as insufficient OSM documentation and tools in Turkey and lack of learning materials in Turkish. Yercizenler and Youth Season NGOs kicked-off the Open Source Volunteering Programme to tackle community needs to build and enlarge OpenStreetMap Turkey community by training volunteers, translating OpenStreetMap related documentation/tools and mapping 120 unmapped towns in Turkey.

The Turkish OpenStreetMap Community has constantly been growing with members from various circles in recent years. It consists of participants such as hobby cartographers, academics, local governments, university student clubs and local NGOs. According to OSMstats, the community, which has an average of forty active mappers daily, has added more than 1.3 million buildings and more than 200,000 km of highways to the OpenStreetMap since 2018. Active OpenStreetMap contributors identify community problems and needs are listed as following; language barrier and the inexperienced local community. According to the EF English Proficiency Index, Turkey ranked in 79 out of 100 countries/regions. Lack of having OSM documentation in Turkish is one of the biggest barriers to growing a mapping community in Turkey. Insufficient OSM documentation in Turkish adversely affects participants of all levels. It causes a lack of technical knowledge about OpenStreetMap data structure among experienced mappers. Even if they make an outstanding amount of contribution, it can cause a decrease in data quality. From the newcomers' point of view, it significantly reduces the pace and motivation of learning. In order to tackle these problems and needs, Yer Çizenler and Youth Season NGOs are teamed up and kicked off the Open Source Volunteering Programme. The main goal of this programme is building and enlarging the OpenStreetMap Turkey community. The objectives of this programme are to train volunteers to get to know more about OpenStreetMap, to translate OpenStreetMap related documentation and tools and to map unmapped towns in Turkey. This programme consists of three phases; training, translation of OSM tools and mapping unmapped towns in Turkey. 8 OpenStreetMap training, four mapathons, three online public experience sharing events and 18 weekly meetings were organized during the programme. Twenty volunteers were selected to be part of the Open Source Volunteering Programme. These volunteers were trained on the following topics; fundamental of OpenStreetMap, OSM data model, OSM tag schema and OSM editing tools such as iD Editor, JOSM, Vespucci, GoMap! and RapiD. With the workshops and hands-on webinars, the participants got familiar with OSM data structure and gained the ability to use OSM editing tools. Additionally, the Open Source Volunteering Programme helped the Turkish OSM community to be connected with international OSM communities to learn from their experiences. As an example, we hosted Geoffrey Kateregga, Community Programs Manager at HOTOSM shared his learning and valuable community growth experience in Uganda as well as in Africa (1). As the second phase of this program, the volunteers translated OSM iD Editor Vespucci, MapRoulette and uMap tools into the Turkish language. In the third phase of the programme, volunteers mapped 120 unmapped towns in Turkey. Unmapped towns are exported from Pascal Neis's Unmapped Places of OpenStreetMap (2). These towns are mapped as points but do not have any roads in a radius of 700m. The MapRoulette and OpenStreetMap iD Editor are used to map unmapped towns. MapRoulette task was used to coordinate collaborative mapping efforts and monitor the progress (3). As a result, 11.838 map changes were made by 15 mappers. There are 10.295 unmapped towns in Turkey. As Yer Çizenler and Youth Season NGO, we're aiming to enlarge the active OpenStreetMap Turkey community and complete a map of unmapped towns in Turkey by the end of 2021. Overall, this talk will highlight key elements of building and enlarging OpenStreetMap communities in developing countries and learnings from Open Source Volunteering Programme. The proof of concept of mapping unmapped towns methodology will be shared with other communities to encourage them to implement similar activities in their local community. * (1) [https://www.youtube.com/watch?v=fD4fvmJ3Wmk](https://www.youtube.com/watch?v=fD4fvmJ3Wmk) * (2) [https://resultmaps.neis-one.org/unmapped#5/47.100/9.800](https://resultmaps.neis-one.org/unmapped#5/47.100/9.800) * (3) [https://maproulette.org/browse/challenges/15901](https://maproulette.org/browse/challenges/15901)

Other people: Oğuzhan Er

Mobile mapping is the process of collecting geospatial data from a mobile vehicle using a 360º camera, laser scanner, GPS/IMU positioning system, and other sensors. This is one of the most time and cost effective methods to collect geospatial data, but the required equipment can be expensive. An alternative approach to mobile mapping brings the opportunity to the OpenStreetMap community to "Map in 360" using a wide range of consumer devices compatible with Mapillary. Instead of capturing a frame of street-level imagery in a single direction, a 360º camera can capture the entire scene simultaneously. 360º street-level imagery provides the OpenStreetMap community a more comprehensive understanding of the map space and more accurate machine-generated map data. In this talk, we will review the workflow of data collection with 360º cameras, look at how to upload 360 street-level images to Mapillary, and compare the impact of different 360º camera models in terms of map data extraction accuracy. OpenStreetMap communities can use this knowledge to coordinate their own street-level imagery collection with 360º cameras to contribute to Mapillary and improve local maps.

Mobile mapping is the process of collecting geospatial data from a mobile vehicle using a 360º camera, laser scanner, GPS/IMU positioning system, and other sensors. Mobile mapping is a time and cost-effective way to conduct asset inventory, monitor road infrastructure, and map street furniture. However, the high cost of acquiring mobile mapping equipment, such as professional 360° cameras and laser scanners, inhibit the ability of local OpenStreetMap communities to leverage mobile mapping technologies. With advice from Mapillary, low-cost devices that are accessible to communities make mobile mapping possible on a wider scale. 360º cameras may be used in different vehicle types: cars, bicycles, quad bikes and also on foot. With a 360º field of view, these cameras capture everything around the sphere. Where flat frame images can limit the map editing workflow, 360º images bring panoramic viewing to OpenStreetMap editors for a more comprehensive understanding of the space and place. The Mapillary photo overlay supports panorama view on OSM ID Editor and JOSM. After images have been captured and uploaded to the web, Mapillary creates a reconstruction of the scene in 3D using an open-source Structure from Motion library (OpenSfM). This helps to improve the image GPS positions, based on the overall consistency of the 3D scene. Next, the objects in the imagery (such as traffic signs and other types of objects) are identified by computer vision. Combining detections of an object in several images and the aforementioned 3D model allows us to calculate the position of the detected object in the world. Overall, this talk will examine the recommended 360º cameras, discuss data collection and benefit of street-level imagery collection with 360º cameras, review strategies for uploading 360º images, comparing different 360º camera models, and highlight how OpenStreetMap communities around the world are using 360º images to improve local maps with Mapillary.

Lightning Talks

## Human in the Loop: Verifying Machine-Generated Data for Better Maps Said Turksever

Machine-generated map data has the potential to considerably accelerate mapping at scale. Combining it with human review helps ensure high data quality. We’ll show how a simple game-based tool helps verify the map data generated by Mapillary’s AI, and how that data helps enhance OpenStreetMap.

## Share the word Ilya Zverev

WeeklyOSM is great, but is it the only channel for following news in OpenStreetMap? For most countries, yes. And that is sad. You can change it for the better: start a blog, record a podcast, tweet something. Here I will share my experience at keeping OSM community informed.

## Enhancing OSM with missing roads Beata Tautan-Jancso

ImproveOSM is a powerful tool that detects and highlights areas in OSM where roads, one-way attributes, and turn restrictions are missing from the map. The first version of the JOSM plugin was released in 2015. Since then, the community has improved almost 200.000 areas with missing roads.

## Community led mapping helping in policy changes Sibabrata Choudhury

Beginning of 2015 a process of community consultation and community led advocacy in the eastern state of India has resulted in several communities developing maps of their plots which has been a breakthrough experience.

## How to create a data annotation process used for navigation Alina Negreanu

In this talk, the Telenav OpenTerra team will present how they built their data annotation team and the processes they developed in order to assure high-quality annotations, on which their AI algorithms heavily rely on. They will talk about good practices to use when building manual datasets and the hurdles they had to overcome in order to reach their quality requirements, having so far reached more than 600 000 annotations.

Other people: Alina Negreanu Sibabrata Choudhury Beata Tautan-Jancso Ilya Zverev