Kimmer VisionAI
A first, clickable look at AI-assisted drill-core logging. Kimmer VisioAI is a pre-release UI prototype: it renders the planned Tray Viewer, Predictions and Model workspace over fictional sample data so exploration teams can react to the workflow before the model ships. Nothing here logs your real core yet, this build does not analyze your core and reads no photos.
A clickable preview of the logging workflow.
Every screen is real UI over a fixed set of sample data. This build does not analyze your data, the workspace exists so geologists can react to the flow before the AI is built.
The full logging workspace, previewed
A three-tab desktop workspace, Tray Viewer, Predictions, and Model, with a left tree over 12 core trays and 3 named model entries. It is a working UI walkthrough of the intended flow; every value shown is from a fixed set of sample data, not from your data.
Core-tray rendering
Renders 12 core boxes as banded lithology sticks with fracture ticks, over contiguous 6 m depth runs. A Litho / Fractures / RQD overlay toggle switches how each tray is coloured. No image input in this build.
Predictions layout
The Predictions view shows a lithology-composition breakdown (Diorite, Andesite, Monzonite, Breccia, Other) plus summary cards for RQD, fracture density and alteration with per-card confidence. These are illustrative constants that stand in for what real model output would populate.
Model card and per-class metrics
The Model tab presents a model card (task: lithology segmentation over 8 classes) and a precision / recall / F1 table for each class. It defines the reporting format the eventual model is designed to fill; the numbers shown are placeholders, and no analysis runs yet.
Segmentation over real core photos (roadmap)
The product is designed around a model that would classify lithology per pixel from imported core-tray photographs and derive RQD and fracture density from that. None of this analysis exists in this build, it is the target the UI is scaffolded for.
Ships through the same signed pipeline
Even as a preview, the app is code-signed and the KimmerHub launcher verifies each download before it runs, identical to Kubaj, MineViz and Satellite. Download entitlement still applies through the hub.
What this preview is scaffolding toward
The demo exists so exploration geologists can react to the workspace before the model is built. The direction it points at is a desktop tool where you import core-tray photos and get back a first-pass log. None of the following is implemented yet, each item below is planned, not shipping.
- Core-photo import and per-pixel lithology segmentation across 8 classes
- Model-derived RQD and fracture-density estimates, geologist-reviewable per tray
- Approve / re-run review loop with export to a report or CSV
- Local, desktop-only analysis, consistent with KimmerLabs' no-cloud stance on deposit data