Feature/analyse v2

python_sdk/docs/specifications/examples/cd_specs_detect_changes_meshes.py

@@ -1,7 +1,7 @@
 import reality_capture.specifications.change_detection as change_detection
 
-cd_inputs = change_detection.ChangeDetectionInputs(model3dA="279db84b-090f-4922-b9a5-7a4fd0a71fcd",
-                                                   model3dB="40af080d-7ata-48c8-974c-610820fe90f2")
-cd_outputs = [change_detection.ChangeDetectionOutputsCreate.OBJECTS3D]
+cd_inputs = change_detection.ChangeDetectionInputs(model3DA="279db84b-090f-4922-b9a5-7a4fd0a71fcd",
+                                                   model3DB="40af080d-7ata-48c8-974c-610820fe90f2")
+cd_outputs = [change_detection.ChangeDetectionOutputsCreate.LOCATIONS3D_AS_GEOJSON]
 cd_options = change_detection.ChangeDetectionOptions()
 cds = change_detection.ChangeDetectionSpecificationsCreate(inputs=cd_inputs, outputs=cd_outputs, options=cd_options)

python_sdk/docs/specifications/examples/clearance_specs.py

@@ -1,6 +1,6 @@
 import reality_capture.specifications.clearance as clearance
 
-clearance_inputs = clearance.ClearanceInputs(model3d="084985b0-71b5-4b02-a788-db261dd0730c",
+clearance_inputs = clearance.ClearanceInputs(model3D="084985b0-71b5-4b02-a788-db261dd0730c",
                                              clearanceFootprint="635f801b-82cc-4477-8d59-f01eb2fea1d9")
 clearance_outputs = [clearance.ClearanceOutputsCreate.OVF_AREAS, clearance.ClearanceOutputsCreate.OVF_LINES]
 clearance_specs =  clearance.ClearanceSpecificationsCreate(inputs=clearance_inputs, outputs=clearance_outputs)

python_sdk/docs/specifications/examples/eval_s3d_specs.py

@@ -2,6 +2,6 @@
 
 eval_s3d_inputs = eval_s3d.EvalS3DInputs(reference="587a14fd-305a-474c-b037-26d4ee8829d9",
                                          prediction="08342927-859c-4563-a4b8-6c6cfb7d5bb3")
-eval_s3d_outputs = [eval_s3d.EvalS3DOutputsCreate.SEGMENTATION3D, eval_s3d.EvalS3DOutputsCreate.SEGMENTED_POINT_CLOUD,
+eval_s3d_outputs = [eval_s3d.EvalS3DOutputsCreate.SEGMENTATION3D, eval_s3d.EvalS3DOutputsCreate.SEGMENTED_MODEL_3D,
                     eval_s3d.EvalS3DOutputsCreate.REPORT]
 eval_s3ds = eval_s3d.EvalS3DSpecificationsCreate(inputs=eval_s3d_inputs, outputs=eval_s3d_outputs)

python_sdk/docs/specifications/examples/s3d_specs_segment_meshes.py

@@ -3,6 +3,6 @@
 s3d_inputs = segmentation3d.Segmentation3DInputs(meshes="401975b7-0c0a-4498-5896-84987921f4bb",
                                                  pointCloudSegmentationDetector="08342927-859c-4563-a4b8-6c6cfb7d5bb3")
 s3d_outputs = [segmentation3d.Segmentation3DOutputsCreate.SEGMENTATION3D,
-               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_POINT_CLOUD]
+               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_MODEL_3D]
 s3d_options = segmentation3d.Segmentation3DOptions()
 s3ds = segmentation3d.Segmentation3DSpecificationsCreate(inputs=s3d_inputs, outputs=s3d_outputs, options=s3d_options)

python_sdk/docs/specifications/examples/s3d_specs_segment_pc.py

@@ -3,6 +3,6 @@
 s3d_inputs = segmentation3d.Segmentation3DInputs(pointClouds="401975b7-0c0a-4498-5896-84987921f4bb",
                                                  pointCloudSegmentationDetector="08342927-859c-4563-a4b8-6c6cfb7d5bb3")
 s3d_outputs = [segmentation3d.Segmentation3DOutputsCreate.SEGMENTATION3D,
-               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_POINT_CLOUD]
+               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_MODEL_3D]
 s3d_options = segmentation3d.Segmentation3DOptions()
 s3ds = segmentation3d.Segmentation3DSpecificationsCreate(inputs=s3d_inputs, outputs=s3d_outputs, options=s3d_options)

python_sdk/docs/specifications/examples/s3d_specs_segment_pc_and_infer_3d_lines.py

@@ -3,7 +3,7 @@
 s3d_inputs = segmentation3d.Segmentation3DInputs(meshes="401975b7-0c0a-4498-5896-84987921f4bb",
                                                  pointCloudSegmentationDetector="08342927-859c-4563-a4b8-6c6cfb7d5bb3")
 s3d_outputs = [segmentation3d.Segmentation3DOutputsCreate.SEGMENTATION3D,
-               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_POINT_CLOUD,
+               segmentation3d.Segmentation3DOutputsCreate.SEGMENTED_MODEL_3D,
                segmentation3d.Segmentation3DOutputsCreate.LINES3D]
 s3d_options = segmentation3d.Segmentation3DOptions()
 s3ds = segmentation3d.Segmentation3DSpecificationsCreate(inputs=s3d_inputs, outputs=s3d_outputs, options=s3d_options)

python_sdk/docs/specifications/examples/training_s3d_default.py

@@ -0,0 +1,14 @@
+import reality_capture.specifications.training as training
+
+training_s3d_inputs = training.TrainingS3DInputs(
+    segmentations3D=["401975b7-0c0a-4498-5896-84987921f4bb"],
+    detectorName="example-detector",
+)
+
+training_s3d_outputs = [
+    training.TrainingS3DOutputsCreate.DETECTOR,
+]
+
+training_s3ds = training.TrainingS3DSpecificationsCreate(
+    inputs=training_s3d_inputs, outputs=training_s3d_outputs
+)

python_sdk/docs/specifications/examples/training_s3d_options.py

@@ -0,0 +1,22 @@
+import reality_capture.specifications.training as training
+
+training_s3d_inputs = training.TrainingS3DInputs(
+    segmentations3D=["401975b7-0c0a-4498-5896-84987921f4bb"],
+    detectorName="example-detector",
+)
+
+training_s3d_outputs = [
+    training.TrainingS3DOutputsCreate.DETECTOR,
+]
+
+training_s3d_options = training.TrainingS3DOptions(
+    epochs=2,
+    spacing=0.2,
+    model=training.Segmentation3DTrainingModel.SPLATNET,
+    features=[training.PointCloudFeature.RGB, training.PointCloudFeature.INTENSITY],
+    versionNumber="1.0",
+)
+
+training_s3ds = training.TrainingS3DSpecificationsCreate(
+    inputs=training_s3d_inputs, outputs=training_s3d_outputs, options=training_s3d_options
+)

python_sdk/docs/specifications/index.rst

@@ -34,6 +34,7 @@ Specifications regroup all the settings used to create jobs with our APIs.
     eval_s3d
     eval_sortho
     clearance
+    training_s3d
 
 
 Modeling
@@ -60,12 +61,14 @@ Analysis
 * :doc:`/specifications/change_detection` will take two point clouds or two meshes to to get 3D regions that capture the changes.
 * :doc:`/specifications/eval_o2d`, :doc:`/specifications/eval_o3d`, :doc:`/specifications/eval_s2d`, :doc:`/specifications/eval_s3d` and :doc:`/specifications/eval_sortho` will compare a prediction to a reference for a specific detection.
 * :doc:`/specifications/clearance` uses a 3D model and a footprint to compute clearance information.
+* :doc:`/specifications/training_s3d` will train an s3d detector from ContextScenes
 
 .. Conversion
 .. ==========
 
 .. * :doc:`/specifications/point_cloud_conversion` will convert point clouds from one format to another.
 
+
 Utilities
 =========
 

python_sdk/docs/specifications/objects2d.rst

@@ -4,7 +4,7 @@ Objects 2D
 
 The *Objects 2D* job uses a photo object detector to detect 2D objects in photos. This job produces a context scene annotated with 2D objects. 
 If the context scene input is oriented, it can turn these 2D objects into 3D objects and produce a context scene annotated with 3D objects. 
-Optionally, these 3D objects that can be exported in another formats, such as 3D Tiles, DGN, SHP, or GeoJSON.
+Optionally, these 3D objects that can be exported in another formats, such as 3D Tiles, SHP, or GeoJSON.
 
 .. contents:: Quick access
    :local:
@@ -36,7 +36,6 @@ This job has three different purposes :
        | *meshes* (optional)
      - | *objects2d*,
        | *objects3d*,
-       | *objects3d_as_dgn* (optional),
        | *objects3d_as_3d_tiles* (optional),
        | *objects3d_as_geojson* (optional),
        | *locations3d_as_shp* (optional),
@@ -52,7 +51,6 @@ This job has three different purposes :
        | *point_clouds* (optional),
        | *meshes* (optional)
      - | *objects3d*,
-       | *objects3d_as_dgn* (optional),
        | *objects3d_as_3d_tiles* (optional),
        | *objects3d_as_geojson* (optional),
        | *locations3d_as_shp* (optional),

python_sdk/docs/specifications/segmentation2d.rst

@@ -2,7 +2,7 @@
 Segmentation 2D
 ===============
 
-The Segmentation 2D job uses a photo segmentation detector to classify pixels in photos. If photos are oriented, it can project the segmentation on meshes or point clouds, and detect 3D lines and 3D polygons. This job produces segmented photos and a context scene pointing to these photos. Optionally, it can also produce context scenes annotated with 3D polygons and 3D lines. Those can be exported in another formats, such as 3D Tiles, DGN, or GeoJSON.
+The Segmentation 2D job uses a photo segmentation detector to classify pixels in photos. If photos are oriented, it can project the segmentation on meshes or point clouds, and detect 3D lines and 3D polygons. This job produces segmented photos and a context scene pointing to these photos. Optionally, it can also produce context scenes annotated with 3D polygons and 3D lines. Those can be exported in another formats, such as 3D Tiles, or GeoJSON.
 
 .. contents:: Quick access
    :local:
@@ -38,11 +38,9 @@ This job has three different purposes :
      - | *segmentation2d*,
        | *segmented_photos*,
        | *lines3d* (optional),
-       | *lines3d_as_dgn* (optional),
        | *lines3d_as_3d_tiles* (optional),
        | *lines3d_as_geojson* (optional),
        | *polygons3d* (optional),
-       | *polygons3d_as_dgn* (optional),
        | *polygons3d_as_3d_tiles* (optional),
        | *polygons3d_as_geojson* (optional)
      - | *min_photos*,
@@ -58,11 +56,9 @@ This job has three different purposes :
        | *segmentation2d*,
        | *point_clouds*/*meshes*
      - | *lines3d* (optional),
-       | *lines3d_as_dgn* (optional),
        | *lines3d_as_3d_tiles* (optional),
        | *lines3d_as_geojson* (optional),
        | *polygons3d* (optional),
-       | *polygons3d_as_dgn* (optional),
        | *polygons3d_as_3d_tiles* (optional),
        | *polygons3d_as_geojson* (optional)
      - | *min_photos*,

python_sdk/docs/specifications/segmentation3d.rst

@@ -2,7 +2,7 @@
 Segmentation 3D
 ===============
 
-The Segmentation 3D job uses a point cloud segmentation detector to classify each point of a point cloud. It produces a segmented point cloud (in OPC format) and a context scene pointing to this point cloud. The segmented point cloud can be exported as LAS, LAZ, POD or PLY. Optionally, it can also produce context scenes annotated with 3D objects, 3D polygons and/or 3D lines. Those can be exported in another format, such as 3D Tiles, SHP, DGN, or GeoJSON.
+The Segmentation 3D job uses a point cloud segmentation detector to classify each point of a point cloud. It produces a segmented point cloud (in OPC format) and a context scene pointing to this point cloud. Optionally, it can also produce context scenes annotated with 3D objects, 3D polygons and/or 3D lines. Those can be exported in another format, such as 3D Tiles, SHP, or GeoJSON.
 
 .. contents:: Quick access
    :local:
@@ -27,11 +27,7 @@ This job has four different purposes :
        | *point_cloud_segmentation_detector*,
        | *clip_polygon* (optional)
      - | *segmentation3d*,
-       | *segmented_point_cloud*,
-       | *segmentation3d_as_las* (optional),
-       | *segmentation3d_as_laz* (optional),
-       | *segmentation3d_as_pod* (optional),
-       | *segmentation3d_as_ply* (optional)
+       | *segmented_model_3d*
      - | *save_confidence*,
        | *crs*
    * - | Segment a collection
@@ -40,11 +36,7 @@ This job has four different purposes :
        | *point_cloud_segmentation_detector*,
        | *clip_polygon* (optional)
      - | *segmentation3d*,
-       | *segmented_point_cloud*,
-       | *segmentation3d_as_las* (optional),
-       | *segmentation3d_as_laz* (optional),
-       | *segmentation3d_as_pod* (optional),
-       | *segmentation3d_as_ply* (optional)
+       | *segmented_model_3d*
      - | *save_confidence*,
        | *crs*
    * - | Segment a collection
@@ -55,23 +47,16 @@ This job has four different purposes :
        | *point_cloud_segmentation_detector*,
        | *clip_polygon* (optional)
      - | *segmentation3d*,
-       | *segmented_point_cloud*,
-       | *segmentation3d_as_las* (optional),
-       | *segmentation3d_as_laz* (optional),
-       | *segmentation3d_as_pod* (optional),
-       | *segmentation3d_as_ply* (optional),
+       | *segmented_model_3d*,
        | *objects3d* (optional),
-       | *objects3d_as_dgn* (optional),
        | *objects3d_as_3d_tiles* (optional),
        | *objects3d_as_geojson* (optional),
        | *locations3d_as_shp* (optional),
        | *locations3d_as_geojson* (optional),
        | *lines3d* (optional),
-       | *lines3d_as_dgn* (optional),
        | *lines3d_as_3d_tiles* (optional),
        | *lines3d_as_geojson* (optional),
        | *polygons3d* (optional),
-       | *polygons3d_as_dgn* (optional),
        | *polygons3d_as_3d_tiles* (optional),
        | *polygons3d_as_geojson* (optional)
      - | *remove_small_components*,
@@ -83,22 +68,15 @@ This job has four different purposes :
        | lines and 3D polygons
      - | *segmentation3d*,
        | *clip_polygon* (optional)
-     - | *segmentation3d_as_las* (optional),
-       | *segmentation3d_as_laz* (optional),
-       | *segmentation3d_as_pod* (optional),
-       | *segmentation3d_as_ply* (optional),
-       | *objects3d* (optional),
-       | *objects3d_as_dgn* (optional),
+     - | *objects3d* (optional),
        | *objects3d_as_3d_tiles* (optional),
        | *objects3d_as_geojson* (optional),
        | *locations3d_as_shp* (optional),
        | *locations3d_as_geojson* (optional),
        | *lines3d* (optional),
-       | *lines3d_as_dgn* (optional),
        | *lines3d_as_3d_tiles* (optional),
        | *lines3d_as_geojson* (optional),
        | *polygons3d* (optional),
-       | *polygons3d_as_dgn* (optional),
        | *polygons3d_as_3d_tiles* (optional),
        | *polygons3d_as_geojson* (optional)
      - | *remove_small_components*,

python_sdk/docs/specifications/segmentation_orthophoto.rst

@@ -2,7 +2,7 @@
 Segmentation Orthophoto
 =======================
 
-The Segmentation Orthophoto job uses an orthophoto segmentation detector to classify pixels in orthophotos. It produces segmented orthophotos and a context pointing to these orthophotos. Depending on the detector, it can detect 2D lines and 2D polygons. Optionally, these 2D lines and 2D polygons can be exported as SHP, DGN or GeoJSON files.
+The Segmentation Orthophoto job uses an orthophoto segmentation detector to classify pixels in orthophotos. It produces segmented orthophotos and a context pointing to these orthophotos. Depending on the detector, it can detect 2D lines and 2D polygons. Optionally, these 2D lines and 2D polygons can be exported as SHP or GeoJSON files.
 
 .. contents:: Quick access
    :local:
@@ -28,7 +28,6 @@ Purposes
        | *polygons2d_as_shp* (optional),
        | *polygons2d_as_geojson* (optional),
        | *lines2d* (optional),
-       | *lines2d_as_dgn* (optional),
        | *lines2d_as_shp* (optional),
        | *lines2d_as_geojson* (optional)
      - |

python_sdk/docs/specifications/training_s3d.rst

@@ -0,0 +1,61 @@
+============
+Training S3D
+============
+
+The *Training S3D* job uses a ContextScene containing annotated pointclouds to train a new S3D detector.
+
+.. contents:: Quick access
+   :local:
+   :depth: 2
+
+Purpose
+=======
+
+  This job has the following purposes:
+
+
+.. list-table::
+   :widths: auto
+   :header-rows: 1
+
+   * - Purpose
+     - Inputs
+     - Possible outputs
+     - Useful options
+   * - Train new detector on a dataset (ContextScene)
+     - | *scene*,
+     - | *detector*,
+     - | *epochs*
+       | *max_train_split*
+
+
+Examples
+========
+
+In this example, we will create a specification for submitting a Training S3D job to produce a new S3D detector from a ContextScene.
+
+.. literalinclude:: examples/training_s3d_new_default.py
+  :language: Python
+
+Classes
+=======
+
+.. currentmodule:: reality_capture.specifications.training
+
+.. autopydantic_model:: TrainingS3DSpecificationsCreate
+
+.. autoclass:: TrainingS3DOutputsCreate
+    :show-inheritance:
+    :members:
+    :undoc-members:
+
+.. autopydantic_model:: TrainingS3DSpecifications
+
+.. autopydantic_model:: TrainingS3DInputs
+    :model-show-json: False
+
+.. autopydantic_model:: TrainingS3DOutputs
+    :model-show-json: False
+
+.. autopydantic_model:: TrainingS3DOptions
+    :model-show-json: False