removed __init__.py in test that pytest didn't

need after all
update manual for the epsg
add a test for epsg

docs/source/tutorials/make_predictions.md

@@ -28,7 +28,8 @@ To show you current inference config, simply add a `--help` flag:
 python run.py task.task_name=predict --help
 ```
 
-Note that `predict.src_las` may be any valid glob pattern (e.g. `/path/to/multiple_files/*.las`), in order to **predict on multiple files successively**.
+Note that `predict.src_las` may be any valid glob pattern (e.g. `/path/to/multiple_files/*.las`), in order to **predict on multiple files successively**.  
+If the lidar file doesn't specify an EPSG in its meatadata, it HAS TO BE be specified with `datamodule.epsg=...`
 
 ## Run inference from sources
 

docs/source/tutorials/prepare_dataset.md

@@ -25,6 +25,8 @@ Under the hood, the path of each LAS file will be reconstructed like this: '{dat
 
 Large input point clouds need to be divided in smaller clouds that can be digested by segmentation models. We found that a receptive field of 50m x 50m was a good balance between context and memory intensity. The division is performed once, to avoid loading large file in memory multiple times during training.
 
+To be able to read the lidar files, an EPSG is needed. If the files don't all specify an EPSG in their metadata, it should be given as a parameter with `datamodule.epsg=...` 
+
 After division, the smaller clouds are preprocessed (i.e. selection of specific LAS dimensions, on-the-fly creation of dimensions) and regrouped into a single HDF5 file whose path is specified via the `datamodule.hdf5_file_path` parameter. 
 
 The HDF5 dataset is created at training time. It should only happens once. Once this is done, you do not need sources anymore, and simply specifying the path to the HDF5 dataset is enough (there is no need for data_dir or split_csv_path parameters anymore).

myria3d/pctl/dataset/utils.py

@@ -9,13 +9,10 @@
 import pandas as pd
 import pdal
 from scipy.spatial import cKDTree
-from pyproj import CRS
 
 SPLIT_TYPE = Union[Literal["train"], Literal["val"], Literal["test"]]
 LAS_PATHS_BY_SPLIT_DICT_TYPE = Dict[SPLIT_TYPE, List[str]]
 
-# commons
-
 
 def find_file_in_dir(data_dir: str, basename: str) -> str:
     """Query files matching a basename in input_data_dir and its subdirectories.
@@ -85,18 +82,20 @@ def get_pdal_reader(las_path: str, epsg: str) -> pdal.Reader.las:
         pdal.Reader.las: reader to use in a pipeline.
 
     """
-    if not epsg:
-        try:
-            crs = CRS.from_string(get_metadata(las_path)['metadata']['readers.las']['srs']['compoundwkt'])
-            epsg = str(crs.to_epsg())
-        except Exception:
-            raise Exception("No EPSG provided, neither in the lidar file or as parameter")
-
-    return pdal.Reader.las(
-        filename=las_path,
-        nosrs=True,
-        override_srs=f"EPSG:{epsg}",
-    )
+
+    if epsg :
+        # if an epsg in provided, force pdal to read the lidar file with it
+        return pdal.Reader.las(
+            filename=las_path,
+            nosrs=True,
+            override_srs=f"EPSG:{epsg}",
+        )
+
+    if 'srs' in get_metadata(las_path)['metadata']['readers.las'] and get_metadata(las_path)['metadata']['readers.las']['srs']['compoundwkt']:
+        # read the lidar file with pdal default
+        return pdal.Reader.las(filename=las_path)
+
+    raise Exception("No EPSG provided, neither in the lidar file or as parameter")
 
 
 def get_pdal_info_metadata(las_path: str) -> Dict:

package_metadata.yaml

@@ -1,4 +1,4 @@
-__version__: "3.5.2"
+__version__: "3.6.0"
 __name__: "myria3d"
 __url__: "https://github.com/IGNF/myria3d"
 __description__: "Deep Learning for the Semantic Segmentation of Aerial Lidar Point Clouds"

tests/conftest.py

@@ -56,6 +56,24 @@ def run_command(command: List[str]):
         pytest.fail(reason=msg)
 
 
+def run_command_with_return_error(command: List[str]):
+    """Default method for executing shell commands with pytest."""
+    msg = None
+    try:
+        sh.python(command)
+    except sh.ErrorReturnCode as e:
+        msg = e.stderr.decode()
+    return msg
+
+
+def run_hydra_decorated_command_with_return_error(command: List[str]):
+    """Default method for executing hydra decorated shell commands with pytest."""
+    hydra_specific_paths = [
+        "hydra.run.dir=" + os.getcwd(),
+    ]
+    return run_command_with_return_error(command + hydra_specific_paths)
+
+
 def run_hydra_decorated_command(command: List[str]):
     """Default method for executing hydra decorated shell commands with pytest."""
     hydra_specific_paths = [

tests/myria3d/test_train_and_predict.py

@@ -11,6 +11,7 @@
 from tests.conftest import (
     make_default_hydra_cfg,
     run_hydra_decorated_command,
+    run_hydra_decorated_command_with_return_error,
     SINGLE_POINT_CLOUD,
     DEFAULT_EPSG,
 )
@@ -90,6 +91,28 @@ def test_predict_as_command(one_epoch_trained_RandLaNet_checkpoint, tmpdir):
     run_hydra_decorated_command(command)
 
 
+def test_command_without_epsg(one_epoch_trained_RandLaNet_checkpoint, tmpdir):
+    """Test running inference by CLI for toy LAS.
+
+    Args:
+        one_epoch_trained_RandLaNet_checkpoint (fixture -> str): path to checkpoint of
+        a RandLa-Net model that was trained for once epoch at start of test session.
+        tmpdir (fixture -> str): temporary directory.
+
+    """
+    # Hydra changes CWD, and therefore absolute paths are preferred
+    abs_path_to_toy_LAS = osp.abspath(TOY_LAS_DATA)
+    command = [
+        "run.py",
+        f"predict.ckpt_path={one_epoch_trained_RandLaNet_checkpoint}",
+        f"predict.src_las={abs_path_to_toy_LAS}",
+        f"predict.output_dir={tmpdir}",
+        "+predict.interpolator.probas_to_save=[building,unclassified]",
+        "task.task_name=predict",
+    ]
+    assert "No EPSG provided, neither in the lidar file or as parameter" in run_hydra_decorated_command_with_return_error(command)
+
+
 def test_predict_on_single_point_cloud(one_epoch_trained_RandLaNet_checkpoint, tmpdir):
     """Test running inference by CLI for cloud with a single point (edge case addressed in V3.4.0)"""
     # Hydra changes CWD, and therefore absolute paths are preferred