I've tried passing a GeoPandas DataFrame several different ways, but each way throws the same exception. This is basically what I'm doing.

// Rust
...
fn my_func(geo: GeoInterface) -> PyResult<()> {
  ...
  Ok(())
}

# Python
gdf: GeoDataFrame = ...

# All of these throw the same exception:
my_func(gdf)
my_func(gdf.geometry)
my_func(gdf['geometry'])
my_func(gdf.__geo_interface__)

Exception:

ValueError: Unsupported geometry type "FeatureCollection"

Any ideas on what I'm doing wrong?

When converting python geometries to geotypes it may be more efficient to check for specific methods of the object first.

For example shapely geometries could be converted to geotypes by using the wkb method when it is available. This would save the overhead of creating the nested python dicts of the geo-interface.