Mapping with RapiD

Posted by jinalfoflia on 7 August 2019 in English (English).

My team at Grab and I got an opportunity to experiment with the RapiD – AI-assisted mapping editor. We found this tool to be useful in mapping missing road networks. It is also helpful in increasing the mapping speed, as:

  • Missing features are already highlighted – helped us focus on the exact location where the feature is missing than going through an entire area.
  • Once selected, they are already traced, thus minimising the tracing efforts and help us better the quality of tracing
  • Help in increasing mapping speed plus give us full control of the data that can be uploaded and the ones that need not be uploaded.

This would be a great tool for the community in SEA as the folks here are new and would be willing to explore this to map the unmapped regions here.

Here are our major takeaways post using the tool:

  • The AI detections are pretty clear and are correctly detecting the roads.
  • The user has the option to accept or reject a detection, clearly establishing that a mapper intervention is necessary even if the AI is pretty good in identifying.
  • It automatically connects the new roads to the preexisting ones or warns when a new segment that we accept has been connected or not, which is great for new mappers to understand and validate.
  • The detections stand out but also doesn’t take up too much space, making it easier to spot the roads without having to disable to rapid assist.
  • The detections already had predefined tags, which makes it quicker for the mapper to add the roads and since the tags are already defined, it eliminates the cases of having or uploading untagged ways.
  • The AI detects the features until the end of that grid and allows the mappers of the other grid to connect the roads, which makes it easy and conflict-free.

The detections have been great. It still allows us to experience the thrill of mapping :) Just makes it easy as it detects places/roads that are missing and through these detections, they are easy to identify, thus making the process faster.

As you can see below, the tool manages to identify and suggest adding roads that align with the imagery. As a mapper and having an eye to making it more precise, I, of course, did add a couple of more nodes but most of my work was done. I managed to map more in lesser time. I loved the fact that even though these are generated, they will be uploaded on OSM if and only if the mapper approves. The quality of the data is in the mapper’s hands.

It also took care of the possible errors which one could have missed. In this case, even though the road is a small extension, the tool suggests the mapper to extend it and trace according to the imagery.

How would it be of use in SEA region:

We would be exploring this tool in dense areas and validate these detections work. We will also be integrating this with our Geo*Stars program and other community activities. This will enable mapping enthusiasts to map in SEA.

Happy Mapping! :)

Login to leave a comment