RobertoCastro391/hillfort-modelling ? reverse-engineered prompt

Build me a project for a dissertation about finding Iron Age hillforts in northern Portugal and Galicia.

I want two workflows in the same repo. One should create a predictive map that ranks the whole study area by how suitable each 5 m cell is for hillforts, using the known sites and terrain and environmental variables. The other should detect hillforts directly from LiDAR relief tiles with object detection and segmentation, then run the detector over tiled imagery.

Also add a small comparison part that measures how long each approach takes over the full area, and lets me test whether running detection only in the most promising zones saves time. Please make it easy to point the scripts to local data with command line paths or simple config files, and use an environment variable for shared figure output paths. The data are huge, so keep the repo code focused on processing and training, not on storing inputs. If you need current docs for the geospatial or deep learning tools, look them up online as you go.