Satellite imagery · analyst workbench
Very large satellite images, rebuilt tile by tile.
Tessera is an interface for analysts to explore Sentinel-2 imagery: deep zoom, before/after comparison, annotation, and review of machine detections. Dense, keyboard first, and built to run on-premise with no network at all.
What's coming
-
01 First
Deep-zoom viewer
Pre-generated tile pyramids over a few areas and dates, with smooth zoom from the whole scene down to the 10 m pixel. An in-house viewer, no map library.
-
02
Before / after
Two dates of the same area under one camera, split by a swipe slider you can drag or drive from the keyboard.
-
03
Annotations
Rectangles and polygons, each with a label and a comment, saved to your own private sandbox. No account needed.
-
04
Detection review
A queue of machine detections to accept, reject or correct by keyboard, with progress tracking. Detections are synthetic in this first version.
-
05
Air-gapped mode
Zero external calls at runtime, proved by an end-to-end test with the network cut, and by an offline install on a machine with no network.
-
06
Case study
The problem, the constraints, the architecture and the measured numbers, in English and French.
Principles
- Air-gapped by design
- No CDN, no external fonts or tiles, no telemetry. This page already loads nothing from anywhere else.
- Frugal dependencies
- Every third-party package is justified in writing. The tile viewer and the charts are built in-house.
- A measured budget
- Bundle size, memory and time to first tile are measured on every change, not guessed.
- Public data, analysis only
- Copernicus Sentinel-2 imagery, prepared on the server. The framing is understanding and situational awareness, never targeting.