You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Density-based clustering with locally-adaptive `ε`, computed in UTM metres so a "20-metre cluster" means the same thing at any latitude. Replaces the prior single-`ε` DBSCAN, which fused downtown into one blob and missed sparser suburban clusters (cf. GraphTrace 2025 urban-analytics benchmarks).
29
+
2.**Planar KDE density surface** (`<city>_heatmap.html` + `<city>_density.geojson`)
30
+
FFT-based kernel density on a metric grid; the folium heatmap is *derived* from the surface rather than from folium's opaque built-in interpolation, and the same surface contours into a GeoJSON layer at the 50/75/90/95 percentiles for researcher-grade work.
The spatial statistic researchers and journalists know from ArcGIS/QGIS "Hot Spot Analysis": per-hex z-score, FDR-adjusted p-value, and a five-class hot/cold classification — the defensible statistical layer per Amnesty's [*Decode Surveillance NYC*](https://decoders.amnesty.org/projects/decode-surveillance) methodology.
33
+
4.**Cameras per road-km** (`<city>_density_metrics.json`)
34
+
The single citable headline number Stanford's [*Surveilling Surveillance*](https://reglab.stanford.edu/projects/surveilling-surveillance/) (2021) made canonical for cross-city comparison. Normalises by the pedestrian network humans actually use, neutralising the park/water/industrial-zone bias of cameras/km². Reuses the routing agent's cached OSMnx graph.
Copy file name to clipboardExpand all lines: pyproject.toml
+6-1Lines changed: 6 additions & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,6 @@
1
1
[project]
2
2
name = "agentic-counter-surveillance"
3
-
version = "2.3.0"
3
+
version = "2.4.0"
4
4
description = "Multi-agent pipeline that scrapes OpenStreetMap surveillance-camera data, enriches it with a local LLM, renders maps and charts, generates an LLM-written city report, and computes low-surveillance walking routes."
0 commit comments