A New Way to Organize Imagery: Introducing Tagging
The ability to add tags or annotations to images is something that we’ve heard about from both our customers and community members. The objective behind this wish is straightforward: tags are helpful for organizing imagery. They will let you mark and then subsequently find images that contain the sort of information you’re interested in. Also, you can share this data with others, whether it’s on a general collaborative basis or with particular people.
For example, a person working with fire safety might use tagging to identify and collect images of fire hydrants in the city. A cyclist might use the feature to find bicycle parking or repair stations. Or someone might mark images that depict construction, so that it’s later easy to follow up those sites with the Time Travel feature (and add imagery where necessary).
Using the feature
The tagging feature has two parts: search and editing. Search will let you find images that have been tagged. You can access it from the Tag Search icon at the top of the Mapillary web app. Once you’re in the tag view, start typing in the search box to get a list of matches. Your chosen tag(s) are represented on the map as yellow dots and you need to zoom into street level to be able to click on them. If you don’t see any dots, try to zoom out or just consider that this tag might now have been added to any image yet.
The latter case will become less and less common as people use the other part of the feature—tag editing. While viewing an image on Mapillary, you can access the Tag Editor via the icon on the right-hand side (it doesn’t matter if the Tag Search mode is switched on or off, but you do need to be signed in).
Start by choosing a tag, using the search box at the upper left corner. Once you’ve picked it, you can opt to add it either to the whole image (e.g. the
Construction tag for a construction scene) or to an object in the image (e.g.
Fire hydrant). In the first case, just click on
Add image tag after selecting the tag from the list. For an object tag, use your mouse to drag a bounding box around the object in the image.
And that’s how easy it is. You can navigate among the images in the Tag Editor just like in the regular view, making it even more convenient to use. For more information and instructions, please refer to our Help Center.
Currently, tags are available from a pre-set selection. Later on, we plan to add the functionality of custom tags. The aim here is to maintain a good balance between customization and systematization—the more people use the same terms to mark the same objects or scenes, the more everyone will benefit from it in the search functionality.
Another intention is to integrate this feature with different tools so that map editing could benefit more smoothly from the ability to find street-level imagery with the relevant content.
In addition, the long-run aim is to use the input from tagging to train our computer vision algorithms for image detections. On the one hand, this will help improve the existing algorithms by creating training sets of images that contain a diverse representation of the same type of object (e.g. parking meters in different cities). On the other hand, this will enable creating training sets for customized detections, so that we can develop new algorithms for objects that are not being detected so far.