Geospatial AI

Geospatial annotation

Label your geospatial and aerial training data with a Human in the Loop

Our expert annotation teams have labeled hundreds of square meters of aerial imagery, with precise segmentation taking up to 8 hours per tile. Using smart automation and robust annotation tools, we are able to annotate thousands of megapixels in a consistent, efficient, and precise way.

Highlight

Client success story

Read how Swiss intelligent drone company Daedalean has used the annotations generated by Humans in the Loop to power the autonomous pilot of the near future

Satellite imagery

Satellite images of entire cities can be segmented with pixel-perfect accuracy for detecting buildings, roads, water bodies, and other land. Other use cases include cloud detection, water bodies monitoring, as well as tracking changes in settlements and land use over time.

Satellite Image pre-segmentation exampleSatellite Image post-segmentation example
Drone Segmentation pre labeling exampleDrone Segmentation post labeling example

Drone imagery

For drones, semantic segmentation can be used for training models for autonomous flight and safe landing. In addition to perception of the environment, drones can be trained to detect and recognize other obstacles and objects of interest, such as other drones.

Inspection

Aerial images can also be labelled with labels such as polygons and bounding boxes. These labels can be used by AI models to inspect the environment and identify or count certain objects, such as houses, trees, vehicles, cattle, electricity poles, and transmission lines.

Aerial inspection pre-labeling exampleAerial inspection post-labeling example
Solar panels pre-labeling exampleSolar panels post-labeling example

Solar panels

Creating AI models that can detect solar panels can be used for large-scale analysis of solar parks. In addition, when applied on infrared imagery, labeling can be used as a time-efficient way to check the condition of solar panels and detect anomalies (short circuits, open circuits, broken glass, and dirt). 

Building footprint

The automatic extraction of building footprints from satellite imagery is vital for a variety of uses such as for city planning and logistical purposes. Precise building polygons on 2D images as well as 3D point cloud annotation of buildings can be used to verify outdated city maps.

Building footprint pre-polygon annotation exampleBuilding footprint post-polygon annotation example
infrared pre-labeling exampleinfrared post-labeling example

Infrared imagery​

By coupling RGB imagery with infrared channels, labeling can be performed to enhance standard detection models in a variety of settings, including night-time imagery, small object detection at sea or in complex settings, or better classification of crops based on their infrared profile.

Natural disasters

Labeled datasets can also be used to train AI models for smart disaster management. These models can be used to efficiently assess the extent of the damage without having to manually assess each incidence of damage, as well as identifying aid, rescue, and evacuation routes.

natural disaster pre-annotation examplenatural disaster post-annotation example

Download our free aerial dataset!

In collaboration with the Mohammed Bin Rashid Space Center Humans in the Loop published an open-access dataset consisting of aerial imagery of Dubai obtained by MBRSC satellites and annotated with pixel-wise semantic segmentation in 6 classes.

Satellite Image post-segmentation example
How would you like your data annotated?

Types of annotation

Bounding Box
Polygon
Keypoint
Segmentation
3-Dimensional
Video

Interested in having a Human in the Loop label your geospatial dataset?
Get in touch with our team at Humans in the Loop and a project manager will help you find the best solution to your computer vision needs!