Skip to content

Commit

Permalink
ENH: Flip the incoming data across the X axis in PeregrineHDF5Reader. (…
Browse files Browse the repository at this point in the history 
…#1086)

PeregrineHDF5Reader: Flip the data across X axis as it is read in.

Signed-off-by: Joey Kleingers <[email protected]>
  • Loading branch information
@joeykleingers
joeykleingers authored Sep 30, 2024
1 parent 8037025 commit f25133d
Showing 1 changed file with 60 additions and 3 deletions.
63 changes: 60 additions & 3 deletions ...g/python/plugins/NXDataAnalysisToolkit/src/NXDataAnalysisToolkit/ReadPeregrineHDF5File.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -769,6 +769,11 @@ def _read_camera_data(self, h5_file_reader: h5py.File, data_structure: nx.DataSt
camera_data_nx[:] = camera_data_h5[slices_subvolume_minmax_z[0]:slices_subvolume_minmax_z[1]+1, slices_subvolume_minmax_y[0]:slices_subvolume_minmax_y[1]+1, slices_subvolume_minmax_x[0]:slices_subvolume_minmax_x[1]+1]
else:
camera_data_nx[:] = camera_data_h5

try:
self._flip_slice_across_x_axis(camera_data_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1031, f"Unable to flip Camera Dataset array '{camera_data_h5_path.as_posix()}' along the X axis: {e}")])

def _read_slice_datasets(self, h5_file_reader: h5py.File, data_structure: nx.DataStructure, filter_args: dict, message_handler: nx.IFilter.MessageHandler, should_cancel: nx.AtomicBoolProxy) -> Result:
read_segmentation_results: bool = filter_args[ReadPeregrineHDF5File.READ_SEGMENTATION_RESULTS_KEY]
Expand Down Expand Up @@ -815,6 +820,11 @@ def _read_slice_datasets(self, h5_file_reader: h5py.File, data_structure: nx.Dat
segmentation_result_nx[:] = segmentation_result_h5[slices_subvolume_minmax_z[0]:slices_subvolume_minmax_z[1] + 1, slices_subvolume_minmax_y[0]:slices_subvolume_minmax_y[1] + 1, slices_subvolume_minmax_x[0]:slices_subvolume_minmax_x[1] + 1]
else:
segmentation_result_nx[:] = segmentation_result_h5

try:
self._flip_slice_across_x_axis(segmentation_result_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1030, f"Unable to flip Segmentation Result array '{segmentation_result_h5_path}' along the X axis: {e}")])

# Read the camera data
if read_camera_data:
Expand All @@ -840,6 +850,11 @@ def _read_slice_datasets(self, h5_file_reader: h5py.File, data_structure: nx.Dat
part_ids_nx[:] = part_ids_h5[slices_subvolume_minmax_z[0]:slices_subvolume_minmax_z[1]+1, slices_subvolume_minmax_y[0]:slices_subvolume_minmax_y[1]+1, slices_subvolume_minmax_x[0]:slices_subvolume_minmax_x[1]+1]
else:
part_ids_nx[:] = part_ids_h5

try:
self._flip_slice_across_x_axis(part_ids_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1032, f"Unable to flip Part Ids array '{ReadPeregrineHDF5File.PART_IDS_H5_PATH}' along the X axis: {e}")])

if read_sample_ids:
message_handler(nx.IFilter.Message(nx.IFilter.Message.Type.Info, 'Reading Sample Ids...'))
Expand All @@ -855,6 +870,11 @@ def _read_slice_datasets(self, h5_file_reader: h5py.File, data_structure: nx.Dat
sample_ids_nx[:] = sample_ids_h5[slices_subvolume_minmax_z[0]:slices_subvolume_minmax_z[1]+1, slices_subvolume_minmax_y[0]:slices_subvolume_minmax_y[1]+1, slices_subvolume_minmax_x[0]:slices_subvolume_minmax_x[1]+1]
else:
sample_ids_nx[:] = sample_ids_h5

try:
self._flip_slice_across_x_axis(sample_ids_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1033, f"Unable to flip Sample Ids array '{ReadPeregrineHDF5File.SAMPLE_IDS_H5_PATH}' along the X axis: {e}")])

return Result()

Expand Down Expand Up @@ -888,6 +908,11 @@ def _read_registered_datasets(self, h5_file_reader: h5py.File, data_structure: n
anomaly_detection_nx[:] = anomaly_detection_h5[registered_data_subvolume_minmax_z[0]:registered_data_subvolume_minmax_z[1]+1, registered_data_subvolume_minmax_y[0]:registered_data_subvolume_minmax_y[1]+1, registered_data_subvolume_minmax_x[0]:registered_data_subvolume_minmax_x[1]+1]
else:
anomaly_detection_nx[:] = anomaly_detection_h5

try:
self._flip_slice_across_x_axis(anomaly_detection_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1034, f"Unable to flip Anomaly Detection array '{ReadPeregrineHDF5File.REGISTERED_ANOMALY_DETECTION_H5_PATH}' along the X axis: {e}")])

if should_cancel:
return Result()
Expand All @@ -907,6 +932,11 @@ def _read_registered_datasets(self, h5_file_reader: h5py.File, data_structure: n
xray_ct_nx[:] = xray_ct_h5[registered_data_subvolume_minmax_z[0]:registered_data_subvolume_minmax_z[1]+1, registered_data_subvolume_minmax_y[0]:registered_data_subvolume_minmax_y[1]+1, registered_data_subvolume_minmax_x[0]:registered_data_subvolume_minmax_x[1]+1]
else:
xray_ct_nx[:] = xray_ct_h5

try:
self._flip_slice_across_x_axis(xray_ct_nx)
except ValueError as e:
return Result(errors=[nx.Error(-1035, f"Unable to flip X-Ray CT array '{ReadPeregrineHDF5File.REGISTERED_XRAY_CT_H5_PATH}' along the X axis: {e}")])

return Result()

Expand Down Expand Up @@ -1020,9 +1050,12 @@ def _read_scan_datasets(self, h5_file_reader: h5py.File, data_structure: nx.Data
return scan_data_result
vertices, edges, tot = scan_data_result.value

# Flip vertices across the X axis
flipped_vertices = self._flip_vertices_across_x_axis(vertices)

# Copy the vertices into the edge geometry
v_end = vertex_tuple_offset + vertices.shape[0]
vertex_list_view[vertex_tuple_offset:v_end, :] = vertices
v_end = vertex_tuple_offset + flipped_vertices.shape[0]
vertex_list_view[vertex_tuple_offset:v_end, :] = flipped_vertices

# Update edges values to match the actual vertices indices
edges += vertex_tuple_offset
Expand All @@ -1040,4 +1073,28 @@ def _read_scan_datasets(self, h5_file_reader: h5py.File, data_structure: nx.Data
vertex_attr_mat.resize_tuples([vertex_tuple_offset])
vertex_list.resize_tuples([vertex_tuple_offset])

return Result()
return Result()

def _flip_slice_across_x_axis(self, slice_arr: np.ndarray):
# Flip slices across the X axis (this is necessary to have the images read in the real coordinate system, not the image coordinate system)
if slice_arr.ndim == 2:
slice_arr[:] = np.flip(slice_arr[:], axis=1)
elif slice_arr.ndim == 3:
for z in range(slice_arr.shape[0]):
slice_arr[:, :, z] = np.flip(slice_arr[:, :, z], axis=1)
else:
raise ValueError("Input array must be either 2D or 3D.")

def _flip_vertices_across_x_axis(self, vertices: np.ndarray) -> np.ndarray:
# Step 1: Compute the centroid
centroid = vertices.mean(axis=0)

# Step 2: Translate vertices to origin
translated = vertices - centroid

# Step 3: Flip Across X-Axis
flipped_translated = translated.copy()
flipped_translated[:, 1] = -flipped_translated[:, 1]

# Step 4: Translate back to original centroid
return flipped_translated + centroid

0 comments on commit f25133d

Please sign in to comment.