GIS Tools Lab — Daud Tasleem
A small but growing collection of focused tools that handle the unglamorous parts of GIS work — semantic metadata, LiDAR strip alignment, tile indexing, dataset quality checks — so analysis can start sooner.
Available tools
Read and write Raster Attribute Tables (RATs) across three backends: GeoTIFF internal TIFF PAM (GDAL), .aux.xml PAM sidecar (lxml), and .vat.dbf dBASE III sidecar. Unified API, auto-detect, pandas interop, and a CLI for RAT inspection.
Ghost-buffered tile streaming for focal raster operations on Cloud-Optimized GeoTIFFs. Applies slope, aspect, hillshade, focal mean, and focal std with zero seam artifacts at tile boundaries. Pure NumPy + rasterio, no extra GDAL dependency.
Streaming Monte Carlo zonal-diff over a baseline land-cover raster and N restoration-scenario rasters. Per-zone delta-ESV with 95 % credible intervals from triangular TEEB coefficient uncertainty, in a single raster sweep.
Detects all interior line-line intersections in a polyline collection and builds a fully planar edge-node graph. Pure NumPy, no PostGIS or ArcGIS required.
Reads GIS datasets, infers semantic metadata, proposes useful spatial questions, and checks whether a custom question is answerable with the available fields.
Explores portable JSON sidecar indexes for LAS/LAZ point clouds, previews bbox query windows, and shows which grid cells a query would read.
Automated boresight calibration and strip adjustment for airborne and UAV LiDAR point clouds. Planar-neighbourhood ICP with Levenberg–Marquardt optimisation, CLI and Python API.
Distance-weighted, spatially-aware evaluation metrics for aerial LiDAR semantic segmentation. Computes per-class Mean Distance Error, spatially stratified IoU by boundary proximity, and hard-point detection. Runs real scipy KD-tree computation in the browser.
Constrained Delaunay triangulation of IFC slab footprints with opening subtraction. Parses IfcSlab and IfcOpeningElement via ifcopenshell, meshes with Shewchuk's Triangle, exports OBJ / PLY / NPZ. 201,810 triangles/s over 200 slabs.
Multi-modal UAV inspection pipeline visualizer for bridges and viaducts. Toggleable RGB, thermal, LiDAR, GPR, and acoustic layers over a real structure, with cross-modal attention fusion and per-defect evidence panels. Aligned to the TU/e STRUCTURE project.
Topology-safe simplification of polygon coverage layers. Detects shared boundary arcs via STRtree, simplifies each arc once (Douglas-Peucker), and reconstructs both neighbours identically — no gaps or slivers introduced. Zoom-aware tolerances via a pixel-error model for web-map pipelines.
Dataset quality checks for CRS, geometry validity, suspicious coordinates, missing context, and publication readiness.