The printed Sketch Map is now ready to be used for your mapping activity. The way you draw on your map determines the detectability of your markings.

To help you in the field, we provide two handouts with best practices that you can print out:

• Infographic
• Mapping tips

The Sketch Map Tool detects shapes (point and polygons) and colors. To make sure your markings are detected:

• Use contrast rich colors - red, blue, green, and black work great
• Use thick and opaque pens (e.g. felt-tip markers)
• Draw closed shapes with continuous lines (not dashed)

Please make sure to not draw over the black-and-white symbols on the map margin.

Tip: Test your pens and the detectibility of your markings before your mapping activity.

Click here for more examples

Recommendations

Line Thickness

Recommended

Use thick felt-tip pens with intense colors.

Not recommended

Thin lines in light colors might not be detected. We generally recommend to test your pens and the detectability of your markings with the Sketch Map Tool before your mapping activity.

Polygon Outlines

Recommended

For polygons, draw their outline as a closed, strong, and continues shape. You do not need to fill the polygons, and the pattern you draw in the middle will not be detected. In general, smaller polygons are detected better than large polygons.

Not recommended

Gaps in the outline could lead to mistakes in the detection.

Large polygons might not be detected properly. If you need to draw large polygons, please consider to zoom out.

If you just draw a line, the marked area of the line will be detected. Nevertheless, we do not recommend to use lines, since the model has not been trained to detect lines and doing so could lead to mistakes in detection of marking.

If you want to draw a point, draw a larger circle to improve the detection. If you want to draw lines, draw strong, non-dashed lines with intense colors.

Hint: If you need the marking as a line or a point shape, you can easily transfer it manually or digitalize it yourself, for example, in QGIS (see Training Exercise 4)

Satellite Base maps

Recommended

Check the darkness of the Satellite base map, especially before printing it in black-and-white.

Make sure you use bright colors and mark the maps especially clearly. Sometimes it might be helpful to use both Sketch Maps with satellite imagery and OSM base maps for the mapping.

Not recommended

Small polygons can be especially difficult to detected.

The gained geodata can be even more valuable for the communities and humanitarian organizations if you create result maps or analyze it further. There are many different analyses possible based on your gained map data.

To open, visualize and work with the downloaded geodata you need some type of GIS. We recommend using QGIS because it is open-source and provides the most flexibility for analyzing and visualizing your data. Alternatively you could also use browser based GIS-Applications like uMap. These browser-based applications are easier to use for beginners, but limited in their functionality. Please also keep in mind data security questions when using Online GIS-Applications for visualizing your data!

For more tips on how to open and analyze the downloaded geodata and how to create a result map in QGIS as well as in uMap, take a look at the following QGIS Training Lectures.

Here you can see an overview on the characteristics of GeoTIFF and GeoJSON and how the results could look like:

GeoTIFF

The GeoTIFF output file consists of a georeferenced "picture" for each of your uploaded Sketch Maps. When you open the GeoTIFF files in a GIS, it should look similar to this example and shows all your sketch map(s) as separate layers.

GeoJSON

If you open the GeoJSON, you can see the detected markings from all uploaded maps as polygons. When you check the attribute table, you can see that for each marking, the name of the map and the detected color are specified. With the help of these attributes, you can style your map or conduct analyses.

Recommendations

• The GeoTIFF file (raster file) can be opened in a GIS software like QGIS but can not be opened in many browser-based GIS-Applications like uMap:
  • It can be used, for example, as a background to combine it with other data or to manually digitize your results yourself in case some markings are not detected correctly.
• The GeoJSON file (vector file) can be opened as well in QGIS as in many browser-based GIS-Applications like uMap:
  • It can be visualized in a clear and concise map in order to use it for reporting (see Training Exercise 4)
  • Based on the vector outputs you can analyze your data and combine it with additional data (e.g. prepare heatmaps, see Training Exercise 5 )
  • In QGIS: You can compare the vector output with the raster file to verify the marking detection. In case features are not detected or are incorrectly detected, you can manipulate your vector file based on the GeoTIFF.
• If the imported data looks strange and does not represent the actual markings of the Sketch Maps, please check your uploaded sketch map files and, if necessary, scan them again or take new pictures. Please review the recommendations in the sections about the data collection, then scan, photograph or the upload above and consult our FAQ section about the problems in the colour detection. If you cannot find your answers there, please contact your Sketch Map Tool focal point in your organization or the Sketch Map Tool or the Sketch Map Tool user support.

• Functionally necessary data storage: We need the original map frame to digitize your Sketch Map. The map frames of the created maps are stored for 12 months, giving you the opportunity to digitize your Sketch Map in those 12 months.

• Functionally necessary data storage: Uploaded images of Sketch Maps are stored for processing (georeferencing and marking detection) until the results are generated.
• The results are stored and available for download for a period of 24 hours, after which they are deleted.

On a voluntary basis, you can enable us to keep your uploaded Sketch Maps beyond processing to further improve the Sketch Map Tool.

If you opt-in by checking the check box you enable us to ...

... keep your uploaded Sketch Maps for future improvements of the Sketch Map Tool such as training our machine learning models which detect your markings on the uploaded images.

The Sketch Map Tool uses a sequence of machine learning models (SAM and YOLO) to detect the markings on the maps. These models need to be trained with a lot of real-world examples so they are able to identity the markings correctly.

Further accuracy improvements can be achieved by incorporating additional real-world cases of marked maps into the models. You can help us here by giving your consent!

For more detailed and technical information take a look at our GitHub and the following figure.

If you opt-out by leaving the checkbox unchecked (default) ...

... your uploaded Sketch Maps will be deleted after the results are generated. The results are stored and available for download for a period of 24 hours, after which they are deleted too.

If you have further questions or any concerns please reach us via E-Mail.

It is also possible to set-up the Sketch Map Tool on your own server. In this case, please get in touch with the Sketch Map Tool team.

Our georeferencing process employs the Binary Robust Invariant Scalable Keypoints (BRISK) method from the OpenCV library for feature matching. It begins by extracting bounding box coordinates and an identification number from a QR code on the uploaded Sketch Map to identify the corresponding unmarked reference map. Using the BRISK algorithm, the marked Sketch Map is aligned with the unmarked map. Later, we use the coordinates of the bounding box to transform the marked map into the world geodetic system (WGS 84), resulting in a georeferenced GeoTIFF file suitable for geospatial analysis.

You can find more details in the paper about the Sketch Map Tool in the paragraph about Image Processing.

The latest version of the Sketch Map Tool leverages advanced Deep Learning models trained to analyze images of Sketch Maps and identify marked objects within them. Through supervised learning on extensive datasets of labeled images, these models acquire the ability to distinguish markings from the background of Sketch Maps by analyzing patterns and features. This enables them to accurately detect and differentiate markings. This predictive capability allows the models to generalize their understanding and accurately identify markings in new and unseen Sketch Maps. The Sketch Map Tool also incorporates conventional Computer Vision techniques to precisely position the Sketch Map within real-world coordinates. Georeferencing images entails detecting and matching specific points, such as the markers in the corners of the Sketch Maps, between the reference and current input image, enabling precise spatial alignment. The algorithm detects all markings as either 'Points' or 'Polygons'. Lines are treated as polygons, and any polygon (whether a line or a shape) below a set size threshold is ultimately classified as a 'Point'.