Satellite and Aerial Image Analysis
Overview
This service provides AI analysis of satellite and drone imagery for monitoring objects, areas, and changes over time. It supports applications such as object counting, repeated area monitoring, and change detection between image acquisitions. The service covers the preparation of input imagery and reference data, adaptation or training of models for the target sensor and geographic context, and delivery of analysis pipelines for operational use. It enables organizations to move from manual interpretation of large image collections to systematic, repeatable remote-sensing workflows. The service uses point localization and counting for object analysis in drone imagery, and bi-temporal change detection for comparing remote-sensing images acquired at different times. The main methodologies used are CeDiRNet for counting and localization, and BTC for change detection, with training or adaptation performed for the target imagery, sensor type, and application context.
Outcome
Deliverables include data-preparation and adaptation workflows, fine-tuned model(s), and inference pipelines for recurring monitoring tasks. Outputs may include object counts, point detections, change maps, per-pixel heatmaps, and summary statistics for reporting and downstream analysis.
Service Card
| Category | AI development and deployment |
| Domain | Earth Observation, Remote Sensing, Geospatial Analytics |
| Vertical | Green Transition |
| Target users | Agrifood Companies, Infrastructure Operators, Environmental Agencies, Urban Planning Bodies, Geospatial Analytics Companies |
| Service type | Model and Pipeline |
| Outputs | Training/adaptation workflow, fine-tuned model(s), and inference results including point detections with counts, change heatmaps/maps, and summary statistics. |
| Requirements | Requires NVIDIA GPU node. Representative aerial or satellite imagery for training. Object counting uses point annotations, while change analysis uses paired images from different time points. Optional domain data can improve adaptation. |
| Limitations | |
| Availability | Q3/2027 |
| Licensing | MIT and CC BY-NC-SA 4.0 |
| Visual content | |
| Visual caption |