Add README.md

README.md

@@ -20,7 +20,7 @@ configs:
 # MALiBU3D
 
 Large-scale **training-ready** airborne LiDAR point clouds over France, with
-land-cover labels, natural-habitat ids, canopy-height (`elevation`), RGB, and
+land-cover labels, natural-habitat axes, canopy-height (`elevation`), RGB, and
 road-network graphs. Built on
 [IGNF/FLAIR-HUB](https://huggingface.co/datasets/IGNF/FLAIR-HUB) 100 m tiles
 plus IGN LiDAR HD. Former working name: Flair3D.
@@ -73,7 +73,7 @@ Each `{roi}.zip` (flat, no wrapping `{roi}/` folder):
 {tile_id}/segment.npy               uint8   (N,)     land cover, Void=15
 {tile_id}/strength.npy              float32 (N,)     LiDAR intensity ~[0, 1]
 {tile_id}/elevation.npy             float32 (N,)     z − DTM (optional)
-{tile_id}/natural_habitat.npy       uint8   (N,)     CarHab ids (optional)
+{tile_id}/natural_habitat.npy       uint8   (N, 4)   ecological axes (optional)
 {tile_id}/forest_2d.npy             uint8   (1, H, W)
 {deptcode}_{roi}_ROADS_graph.gpkg    optional, EPSG:2154
 ```
@@ -106,14 +106,22 @@ global (`labels.json`).
 
 See `labels.json`. Land cover (`segment.npy`): 15 train classes + Void=15.
 
-Natural habitat (`natural_habitat.npy`): 44 CarHab ids, Void=43. Four ecological
-axes are **LUTs** in `labels.json` (lossy). Remap with:
+Natural habitat (`natural_habitat.npy`): **`(N, 4)` uint8**, already remapped
+from CarHab. Column order is `labels.json` → `natural_habitat.columns`:
+
+| col | key | classes | Void |
+| --- | --- | --- | --- |
+| 0 | `nathab_habitat_type` | Open, Forest, Mineral, Aquatic | 4 |
+| 1 | `nathab_moisture_regime` | Humide, Mesique, Sec | 3 |
+| 2 | `nathab_soil_chemistry` | Acidic, Alkaline | 2 |
+| 3 | `nathab_bioclimatic_zone` | Temperate, Mediterranean, Alpine | 3 |
 
 ```python
 import json, numpy as np
 labels = json.load(open("labels.json"))
-lut = np.array(labels["natural_habitat"]["axes"]["by_moisture_regime"]["lut"])
-axis = lut[raw_ids]   # raw_ids = natural_habitat.npy
+nh = np.load("natural_habitat.npy")          # (N, 4)
+moisture = nh[:, 1]                          # Void = 3
+# Pointcept: no LUT. Assign column i to task labels["natural_habitat"]["columns"][i].
 ```
 
 ## `forest_2d` → points
@@ -149,6 +157,6 @@ Derived semantic labels and graphs are produced by this project.
 
 ## Loader notes
 
-On-disk `segment` / `natural_habitat` are **uint8**. Cast if a loader asserts
-another integer dtype. Per-point `forest.npy`, `land_use.npy`, and `network.npy`
-are absent.
+On-disk `segment` is **uint8** `(N,)`. `natural_habitat` is **uint8** `(N, 4)`.
+Cast if a loader asserts another integer dtype. Per-point `forest.npy`,
+`land_use.npy`, and `network.npy` are absent.