docs: update README

README.md

@@ -1,3 +1,52 @@
----
-license: gpl-3.0
----
+---
+license: gpl-3.0
+---
+
+# detectree weights for Belem, Brazil
+
+## Installation
+
+This requrires detectree>=0.8.0, you can install it using pip/uv:
+
+```
+pip install "detectree>=0.8.0"
+```
+
+or conda/mamba:
+
+```
+conda install -c conda-forge "detectree>=0.8.0"
+```
+
+## Example usage
+
+Run from the command line, e.g., to predict for a single image
+
+```
+detectree predict-img <path-to-input-image> --output-filepath <path-to-output-image> --hf-hub-repo-id martibosch/detectree-belem
+```
+
+or to predict for multiple images:
+
+```
+detectree predict-imgs <path-to-output-dir> --hf-hub-repo-id martibosch/detectree-belem --img-dir <path-to-input-dir>
+```
+
+where `<path-to-input-dir>` is the path to the folder containing the `.tif` files. If the image files have another extension, e.g., jpeg, provide the `--img-filename-pattern` option as in:
+
+```
+detectree predict-imgs <path-to-output-dir> \
+    --hf-hub-repo-id martibosch/detectree-belem \
+    --img-dir <path-to-input-dir> \
+    --img-filename-pattern "*.jpeg"
+```
+
+Additionally, you can customize the postprocessing (and many other) parameters (see the [background](https://github.com/martibosch/detectree-examples/blob/main/notebooks/background.ipynb) notebook for more details) as in:
+
+```
+detectree predict-imgs <path-to-output-dir> \
+    --hf-hub-repo-id martibosch/detectree-belem \
+    --img-dir <path-to-input-dir> \
+    --img-filename-pattern "*.jpeg" \
+    --refine-beta 10000
+```