A showcase of the SHaDE Lab's published human-thermal-exposure tools: papers, live tools, interactive demos and open-source example code.
SHaDE Lab · Arizona State University
Status: private review copy. The repository is shared inside the SHaDE-Lab organization so the lab can look it over before it is made public. Everything here is intended to be open source (see License).
Mean radiant temperature (MRT) is the radiation a person absorbs from sun, sky and surrounding surfaces. It often drives outdoor heat stress more than air temperature does. The three projects below estimate it from three kinds of data.
| Project | What it does | Live tool | Paper | Example code |
|---|---|---|---|---|
| Cool Routes | Walking routes on the ASU campus that trade a little distance for much less radiant heat, from hourly SOLWEIG MRT forecasts | coolroutes.sensableheatscapes.com | Buo et al., Building and Environment (in press), doi:10.1016/j.buildenv.2026.114622 | codebases/cool-routes |
| VisionMRT | Physics-based MRT, UTCI and PET from a 360° image: segmentation, depth, shadow casting, surface energy balance | shadelab.asu.edu/VisionMRT | Middel et al., Sustainable Cities and Society 151:108016 (2026), doi:10.1016/j.scs.2026.108016 | codebases/visionmrt |
| WebMRT | Machine-learning prediction of summertime MRT from meteorology and the surroundings (LightGBM, trained on MaRTy field data) | shadelab.asu.edu/WebMRT | AlKhaled et al., Sustainable Cities and Society 115:105861 (2024), doi:10.1016/j.scs.2024.105861 | codebases/webmrt |
| PanoMRT | MRT, UTCI and PET from a thermal 360° panorama and weather data: surface temperatures measured per pixel, six-direction radiation budget (no web tool) | none | Middel et al., Science of The Total Environment 859:160301 (2023), doi:10.1016/j.scitotenv.2022.160301 | codebases/panomrt |
The website also lists the related image-based modeling paper, Shaeri, AlKhaled & Middel (2025), A Multimodal Physics-Informed Neural Network Approach for Mean Radiant Temperature Modeling, ICCV Workshops, IEEE Xplore.
The two instruments behind the lab's field data, each with an interactive 3D model you can rotate and take apart:
| Tool | What it is | Paper | Code |
|---|---|---|---|
| MaRTy | Mobile human-biometeorological station: six-directional radiation (twelve sensors), air temperature, humidity, wind and GPS on a cart | Middel & Krayenhoff, Science of The Total Environment 687:137–151 (2019), doi:10.1016/j.scitotenv.2019.06.085 | codebases/MaRTy (CERN-OHL-P-2.0) |
| MaRTiny | Low-cost IoT device with a globe thermometer and embedded computer vision that counts people in sun and in shade | Kulkarni et al., Frontiers in Environmental Science 10:866240 (2022), doi:10.3389/fenvs.2022.866240 | codebases/MaRTiny (CERN-OHL-P-2.0) |
The 3D models are drawn in code (three.js) from photographs of the devices and the sensor tables in the papers. They are illustrative: proportions are approximate and some mounting details are simplified.
index.html the showcase website (static, no build step)
assets/ styles, scripts and the single place to edit links (PORTAL block in assets/site.js)
demos/ interactive demos that run in the browser
coolroutes/ real ASU campus data and a JavaScript router with the same cost function as the Python code
visionmrt/ precomputed all-day outputs of the released code on two sample sites
webmrt/ the exported LightGBM sample model running in the browser
_3d/ marty/ martiny/ the shared three.js viewer and the MaRTy and MaRTiny 3D models
codebases/ curated, tested example code for each project (each folder has its own README)
cool-routes/ visionmrt/ webmrt/ panomrt/ MaRTy/ MaRTiny/ (MaRTy: logger program and field guide; MaRTiny: Arduino and ESP32-CAM firmware, image server, MongoDB tools and model notes)
patches/ proposed change to the Cool Routes frontend (URL parameters for start and destination)
LICENSE NOTICE Apache-2.0 and third-party notices
Each codebase is a curated release of the published method on small sample data, with tests. It is not the full web application and it does not reproduce each paper's complete validation. Each README lists what is included, what is deliberately left out, the provenance of the sample data and the known limitations.
The demos load their data with fetch, so serve the folder over HTTP (opening index.html from disk will not work):
python -m http.server 8080
# open http://localhost:8080The page opens in dark mode by default; the ◐ button switches themes and the browser remembers the choice.
The 3D viewers load three.js from a pinned jsDelivr URL (see the import map in index.html), so they need an internet connection and a
browser with WebGL.
Each codebase is self-contained (Python 3.10+).
cd codebases/cool-routes && pip install -r requirements.txt && python -m pytest # 21 tests, about 15 s
cd codebases/webmrt && pip install -r requirements.txt && python -m pytest # 17 tests, about 5 s
cd codebases/visionmrt && pip install -r requirements.txt && python -m pytest # 41 tests, about 4 min
cd codebases/panomrt && pip install -r requirements.txt && python -m pytest # 51 tests, about 3 min (-m "not slow": 15 s)The in-browser demos are checked against the Python results (needs Node 18+):
node demos/coolroutes/parity.test.mjs # JS router vs. Python golden routes
node demos/webmrt/parity.test.mjs # JS tree evaluator and solar position vs. Python / pvlibEach codebase has a run_demo.py that regenerates its demo assets under demos/.
All project links and release status live in one block at the top of assets/site.js (window.PORTAL):
the live-tool URL, paper URL, code URL and whether the live tool can be embedded.
- Embedding the live tools. Cool Routes is embedded in the page.
shadelab.asu.educurrently sendsX-Frame-Options: DENY, so WebMRT and VisionMRT show a launch card. If the host allows this site (Content-Security-Policy: frame-ancestors), set theirembedflag totrue. - Start and destination in the embedded Cool Routes.
patches/holds a small proposed change to the Cool Routes frontend that reads?start=COOR&end=CRTVC. It is a proposal only and has not been applied to any lab repository. - Hosting. The site is plain static files (
.nojekyllis included), so it can be served with GitHub Pages once the repository is public, or on any static host.
Sample data are included for demonstration and documented in each codebase README. In short: the Cool Routes walking network derives from OpenStreetMap (© OpenStreetMap contributors, ODbL); the VisionMRT sample weather is from Open-Meteo (CC BY 4.0) and its sites ship as segmentation and depth maps only (no source photographs); the WebMRT sample is a seeded subset of the lab's MaRTy measurements. The SOLWEIG model code used to produce the Cool Routes MRT rasters derives from UMEP (GPL-3.0) and is not included. See NOTICE.
If you use a tool or its code, please cite the paper for that tool (BibTeX is on the website under Publications, and each codebase has
a CITATION.cff).
Code is released under the Apache License 2.0, except codebases/MaRTy and codebases/MaRTiny, which are released under the
CERN Open Hardware Licence v2, Permissive (CERN-OHL-P-2.0) because they describe hardware instruments.
Third-party material keeps its own license (see NOTICE).
The SHaDE Lab, Arizona State University.