pedroivospatial/pis-spatial-intelligence-lab
PIS Spatial Intelligence Lab
An English-first, privacy-safe research demonstrator from Pedro Ivo Spatial.
Current hosted mode: static and pre-computed
The Hugging Face surface is intentionally a Static Space. It presents three deterministic synthetic scenes and three fixed relative-depth previews calculated before publication. Selecting a scene does not run a model, submit data or start a remote inference request. See `STATIC_SPACE.md`.
The original Gradio implementation remains in app.py with its deterministic scene logic in lab_core.py. Its reviewed runtime metadata is preserved here for future local use or an explicitly approved migration:
sdk_version: 6.26.0app_file: app.py
The pre-computed experiment applies the Apache-2.0-metadata `depth-anything/Depth-Anything-V2-Small-hf` model at revision 5426e4f0f36572d16453bbda7a8389317b1bef99 to procedurally generated indoor scenes. The demo accepts no uploads and contains no client captures, panoramas, point clouds, plans, E57 files, personal data, or real locations.
What it demonstrates
- relative monocular depth estimation;
- a visual depth layer for spatial-capture research;
- simple, explicitly experimental depth-signal diagnostics;
- a safe public surface for PIS research without exposing customer material.
What it does not do
- It does not produce metric measurements.
- It is not a floor plan, survey, safety inspection, or digital twin.
- Its diagnostic score is not a quality guarantee.
- It is not a replacement for professional reality capture or Realsee delivery.
Data boundary
Only deterministic synthetic scenes generated by lab_core.py are represented. The interface intentionally has no file-upload, camera, microphone, URL-fetch, or free-text data input. See `DATA_POLICY.md`.
Commercial boundary
The public experiment uses an Apache-2.0 model candidate. Research-only models such as MapAnything and SpatialLM are tracked separately and are not executed by this Space. Every future model must pass a dependency and license review before commercial use.
Review the Static Space locally
python -m http.server 8765Open http://127.0.0.1:8765/.
Run the preserved Gradio laboratory locally
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python app.pyThe local Gradio path downloads the model from its official Hugging Face repository on first use. No model weights are included in this repository and the hosted Static Space does not execute Python or model inference.
About Pedro Ivo Spatial
Pedro Ivo Spatial provides professional spatial capture, hosted 360 experiences, floor-plan and point-cloud add-ons, and reality-capture services in the Netherlands and selected European locations.