Just maps v2

TL;DR

Intro

Coming from this post

https://www.youtube.com/watch?v=4R4xRH-Nyac

https://www.youtube.com/watch?v=uPM2gNSWX9o&list=PLAxJ4-o7ZoPcfLJ0w7k-woHJXkbjSiKb3

https://www.youtube.com/results?search_query=f4map

This is a good chance:

TrackDays and GoPro Telemetry

GPS Telemetry from a GoPro | Post

GoPro MetaData Magic 101

OA5 Firmware and comparison with GoPro

1. To have a look to French Real estate again:

Go to the details of DFV geo exploration

2. To check how household income is evolving in spain as a recap of this post

3. To use PyRouteTracker against more karting or trackday data:

4. Just imagine how a Drone x GPS could be!

5. Phyphox take off data: remember that you can configure replicable experiments via QR!

cd ./poc/airplane-phyphox

6. To think how the solar rays simulations of a building need more context of their surrounding:

#sudo snap install blender --classic --channel=5.1/stable
#cd ./poc/building-to-blender
#claude --dangerously-skip-permissions
cd ./poc/building-geo-to-blender

Go to the details

7. Matching an action cam video to a GPX file: the physics of coasting behind an ebike

SelfHosted GPX

https://github.com/gpxstudio/gpx.studio

MIT | the online GPX file editor

https://gpx.studio/app#10.96/42.9481/-0.2867/0/70

1. GeoLibre - https://github.com/opengeos/GeoLibre that you can find https://geolibre.app/

2. Dawarich - with Android app

You have several integrations http://localhost:3333/settings/integrations like with Velomate

https://github.com/JAlcocerT/Home-Lab/tree/main/dawarich

Make sure to have proper Application hosts in the .env

on a fresh self-hosted install you should be able to sign in with the seeded default account:

• Username: demo@dawarich.app
• Password: safepassword (github.com (https://github.com/Freika/dawarich))

For Cloudflare exposure, the important bit is that the app container must join the shared tunnel network, and Rails must allow the public hostname in APPLICATION_HOSTS.

• If you expose the same service both locally and through Cloudflare, the local port mapping can stay in place. The tunnel just adds another path in.
• The tunnel side does not need a separate reverse proxy here; your existing cloudflared container can route directly to dawarich_app:3000.
• Browser issues can look like app failures, but in this case the service was fine; the problem was the client path and DNS/IPv6 timing.

Go to CF and add dawarich_app:3000

3. More Selfhosted GPX like https://github.com/tess1o/geopulse

4. Geolibre

5. Owntracks

6. https://github.com/traccar/traccar

7. GeoPulse

8. Others:

https://github.com/mendhak/gpslogger/

https://github.com/dietrichmax/colota

https://github.com/dawarich-app/atlas

Mobile GPX Apps

https://apps.apple.com/us/app/open-gpx-tracker/id984503772

2. Dawarich https://play.google.com/store/apps/details?id=com.zeitflow.dawarich&pli=1

GoPro GPS

Building around GeoData

French DFV Prices x Geo

python scripts/build_transaction_map.py
start maps\dvf_eaux_bonnes_2025_transactions.html

python scripts/build_france_transaction_map.py --max-mutations 1000
#  python scripts/build_france_transaction_map.py --department 64
maps\dvf_france_2025_transactions_preview.html

#start maps\dvf_france_2025_transactions_preview.html
start maps\dvf_combined_lyon_dept64_2025_transactions_preview.html

I put together a quick video here around those

python -m py_compile scripts\build_price_video.py

See where the sold houses are and how much sun they are getting in December, by combining: 22.8% of the inhabited corridor gets zero direct sun in December.

Roughly nine steps, and the ordering has one non-obvious dependency: the prices decide how big the DEM has to be, so you can’t fetch terrain first.

Pass 1 — the transaction study

1. Probe the source before writing anything. curl -sI on files.data.gouv.fr/geo-dvf/latest/csv/{year}/communes/64/64204.csv to find out what actually exists. Result: 2021–2025 only (latest is a rolling five-year window; 2014–2020 all 404). That single check decided the sample size for everything downstream.

2. Look at raw rows before designing the €/m² rule. Downloaded the five files (~180 KB total) and dumped a few multi-row mutations. That’s where the rule came from: Dépendance rows carry no surface_reelle_bati at all, so Appartement; Dépendance is one flat sold with its cellar — keep it, and use the flat’s surface as the denominator. Had I written the ingest first I’d have excluded those and thrown away most of the sample.

3. Ingest → fetch_geodvf.py. Group by id_mutation (the official key — no date+value+address heuristic needed), apply the pricing rule, log a reason for every rejection. → 485 mutations, 481 geocoded, 397 priced, median 1686 €/m².

4. Now size the terrain from the prices. Bbox of the transaction coordinates: 3.1 × 6.9 km, furthest point 3.8 km from centre. Since the horizon march needs 30 km, the DEM radius must be 30 + 3.8 ≈ 34 km. That’s the dependency — terrain extent is a function of where the sales are.

5. DEM fetch → fetch_terrain.py (copied verbatim from building-geo-to-blender). 110 terrarium tiles, 34 MB, 80 s, no API key, no GDAL.

6. Horizon per parcel → sun_by_point.py. The one piece of genuinely new geometry: the sibling project marches rays from the centre of its DEM, this marches from an arbitrary offset inside a shared one. Deduped to 117 distinct coordinates rather than 481 mutations, because DVF geocodes to the parcel. 17 s.

7. Join and test → join_analyze.py. Correlations, cluster-robust regressions, the faceted scatter, the Leaflet map. → the null result, which I then blamed on a +0.92 elevation/sun collinearity.

geo-DVF csv ──> mutations.csv ──┬──> bbox ──> DEM ──> horizons ──> sun_by_point.csv
                                │                                        │
                                └────────────────  join  ────────────────┘
                                                     ↓
                                        findings.json, chart, map

Pass 2 — the raster that overturned step 7’s diagnosis

8. Same DEM, grid instead of points → build_sun_raster.py. 4 515 cells, 5 min 20 s. Then — before trusting a single number — cross-check: sample the raster at the 117 parcel coordinates and compare against pass 1. r = 0.999, mean absolute difference 1.5 h, signed difference −0.05 h.

9. Bin by elevation → analyze_bands.py. This is where the reframing happened: over the terrain the collinearity is only +0.24, versus +0.74 over sold parcels and +0.92 over sales. So the matched-elevation test became possible, and it made the null stronger.

10. Composite → composite_map.py. Blend the raster onto real OSM tiles. Nominally an illustration; actually the georeferencing check — the pale corridor lands exactly on the Valentin and the D918.

The shape underneath it

Three habits did most of the work, and they’re all lifted from your existing projects rather than invented:

• Inspect the data before writing the parser. Steps 1 and 2 each changed the design.
• Two independent paths, then check they agree. Your building-geo-to-blender cross-checks its analytic horizon against a Cycles render (38.4% vs 38.9%); here the raster cross-checks the point run, and the composite cross-checks the georeferencing. Every expensive claim has a cheap second opinion.
• Refuse rather than degrade. Both the point script and the raster script exit with the march radius that would fit instead of letting rays edge-clamp — because edge-clamping silently flattens distant ridges into a sunnier answer and leaves no trace in the output.

Stack: Python stdlib for the data plumbing, numpy for the marching, scipy for the statistics, Pillow for the rasters, matplotlib for the charts, Leaflet for the maps. No database, no GDAL, no npm.

Solar Rays x Buildings in a Geolocation

For an off-grid or heat-pump build, annual totals are the wrong statistic and December is the whole design constraint.

https://jalcocert.github.io/JAlcocerT/data-driven-insulation-evaluation/#the-sun-is-interesting

Blender x GIS

• https://github.com/domlysz/BlenderGIS

And with blenderGIS you can create interesting simulations:

Video x GPX matching

Coasting behind an ebike

The amount of power (assistance) an e-bike provides depends on the motor rating, the jurisdiction/legal limits, and the difference between continuous vs. peak power.

• https://github.com/JAlcocerT/hermesagent/tree/tinker/hermesagent/mbsd/bike-practicas-cdm

1. Typical Power Ratings (Nominal / Continuous)

• 250 Watts (Standard / EU, UK, Australia): In Europe, the UK, and Australia, the legal maximum for a standard pedal-assist bike (pedelec) is 250 W continuous rated power.
• 500 W to 750 Watts (US & Canada): In the United States and Canada, legal e-bikes (Class 1, 2, and 3) typically allow up to 750 W of nominal motor power.
• 1,000 W+ (Off-Road / Speed Pedelecs / DIY): High-power cargo bikes, speed pedelecs, or off-road e-bikes range from 1,000 W up to 1,500 W+.

2. Nominal (Continuous) vs. Peak Power

Advertised motor wattage usually falls into two categories:

3. Human Power vs. E-Bike Power

To put e-bike assistance in perspective relative to human effort:

• Average Human Cyclist: Produces around 100 W to 150 W of sustained pedaling power.
• Fit / Amateur Cyclist: Produces around 200 W to 250 W.
• Pro Cyclist: Sustains 350 W to 450 W during intense efforts.

You can imagine how hard was to cross from UK to FR by a bike powered plane!

For a non-professional cyclist, power output depends heavily on fitness level, body weight, and duration.

Cycling power is typically measured at FTP (Functional Threshold Power—the maximum average wattage you can hold for roughly 1 hour) or measured in Watts per kilogram ($\text{W/kg}$).

Sustained Power (1-Hour Continuous Effort)

For an average adult male weighing around $75\text{ kg } (165\text{ lbs})$:

Short-Burst / Sprint Power

If you ask a non-professional to stomp on the pedals as hard as possible for a brief burst:

• 5–10 Second Max Sprint: An average non-pro can burst $500\text{ to }900\text{ Watts}$. A strong amateur sprinter can peak around $1,000\text{ to }1,200\text{ Watts}$.
• 1-Minute Hard Effort: A healthy non-pro can hold $300\text{ to }450\text{ Watts}$ before severe fatigue and acid buildup set in.

Bike Reverse Engineering Experiment

I remember that we made an experiment during the studies where we logged bike sensor data

This can be useful to estimate the power i applied to the bike

git clone 
cd ./

Conclusions

Not sure about you, but im doing:

flowchart LR
    %% --- Styles ---
    classDef free fill:#E8F5E9,stroke:#2E7D32,stroke-width:2px,color:#1B5E20;
    classDef low fill:#FFF9C4,stroke:#FBC02D,stroke-width:2px,color:#FBC02D;
    classDef mid fill:#FFE0B2,stroke:#F57C00,stroke-width:2px,color:#F57C00;
    classDef high fill:#FFCDD2,stroke:#C62828,stroke-width:2px,color:#C62828;
    classDef bridge fill:#E3F2FD,stroke:#1565C0,stroke-width:3px,color:#0D47A1;

    %% --- Nodes ---
    L0("Free Content
( DIY = $0)"):::free
    L1("Web Audits 🛡️
(Reveals Problem )"):::free
    L11("Tech Blog/Youtube"):::free
    L12("ebooks"):::free
    L13("mbsd framework OSS"):::free
    L14("OSS guides"):::free

    L3("Done With You
(Trade $$ for knowledge)"):::mid
    L4("Done For You
(Trade $$$ for outcomes)"):::high
    L44("GenBI
Shopify PoC"):::bridge
    L45("Real Estate
Funnel Bot"):::bridge
    L46("Energy Solutions
HVAC"):::bridge
    L47("IoT Solutions
Crops"):::bridge
    L48("Weddings
Photo QR"):::bridge

    %% --- Connections ---
    L0 --> L1
    L1 --> L3
    L12 --> L3
    L13 -->|MultiBodySystemsDynamicscom| L3
    L14 -->|FOSS Engineer| L3
    L0 --> L11
    L0 --> L12
    L0 --> L13
    L0 --> L14
    L3 --> L4
    L4 -->|Productized Service| L44
    L4 -->|Productized Service| L45
    L4 -->|Productized Service| L46
    L4 -->|Productized Service| L47
    L4 -->|Productized Service| L48

FAQ

Finding interesting housing data

1. For Spain

• https://penotariado.com/inmobiliario/en/housing-price-finder

You can correlate with household income

2. For FR:

3. For DK:

4. For PL:

GPS Tracker

https://www.traccar.org/docker/ https://github.com/traccar/traccar-docker