v1 dataset release

code/python/analysis/dataset_generate_image_metadata.py

@@ -0,0 +1,128 @@
+#
+# For licensing see accompanying LICENSE.txt file.
+# Copyright (C) 2020 Apple Inc. All Rights Reserved.
+#
+
+from pylab import *
+
+import argparse
+import os
+import pandas as pd
+
+import path_utils
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--analysis_dir", required=True)
+args = parser.parse_args()
+
+assert os.path.exists(args.analysis_dir)
+
+
+
+print("[HYPERSIM: DATASET_GENERATE_IMAGE_METADATA] Begin...")
+
+
+
+num_images_per_camera_trajectory = 100
+
+metadata_images_flagged_txt_file = os.path.join(args.analysis_dir, "metadata_images_flagged.txt")
+metadata_images_csv_file         = os.path.join(args.analysis_dir, "metadata_images.csv")
+
+metadata_camera_trajectories_csv_file = os.path.join(args.analysis_dir, "metadata_camera_trajectories.csv")
+df_camera_trajectories = pd.read_csv(metadata_camera_trajectories_csv_file).rename_axis("camera_trajectory_id")
+camera_trajectories = df_camera_trajectories.to_records()
+
+
+
+# initialize dict of lists for frames to exclude
+frames_ids_excluded_flagged = {}
+
+for c in camera_trajectories:
+
+    animation_name = c["Animation"]
+
+    scene_name  = animation_name[0:10]
+    camera_name = animation_name[11:17]
+
+    assert scene_name.startswith("ai_")
+    assert camera_name.startswith("cam_")
+
+    frames_ids_excluded_flagged[(scene_name, camera_name)] = []
+
+
+
+# fill lists of frames to exclude
+scene_name_current  = None
+camera_name_current = None
+frame_id_current    = None
+
+for line in open(metadata_images_flagged_txt_file, "r"):
+
+    line = line.strip()
+
+    assert line == "" or line.startswith("ai_") or line.startswith("scene_cam_") or line.startswith("frame.")
+
+    if line == "":
+        continue
+    if line.startswith("ai_"):
+        scene_name_current = line[0:10]
+    if line.startswith("scene_cam_"):
+        camera_name_current = line[6:12]
+    if line.startswith("frame."):
+        assert scene_name_current is not None and camera_name_current is not None
+        frame_id_current = int(line[6:10])
+        frames_ids_excluded_flagged[(scene_name_current, camera_name_current)].append(frame_id_current)
+
+
+
+# create dataframe
+df_columns = ["scene_name", "camera_name", "frame_id", "included_in_public_release", "exclude_reason"]
+df = pd.DataFrame(columns=df_columns)
+
+for c in camera_trajectories:
+
+    animation_name = c["Animation"]
+
+    scene_name  = animation_name[0:10]
+    camera_name = animation_name[11:17]
+
+    assert scene_name.startswith("ai_")
+    assert camera_name.startswith("cam_")
+
+    print("[HYPERSIM: DATASET_GENERATE_IMAGE_METADATA] Processing scene: " + scene_name)
+
+    scene_names  = [ scene_name for i in range(num_images_per_camera_trajectory) ]
+    camera_names = [ camera_name for i in range(num_images_per_camera_trajectory) ]
+    frame_ids    = range(num_images_per_camera_trajectory)
+
+    if c["Scene type"] == "OUTSIDE VIEWING AREA (BAD INITIALIZATION)":
+        included_in_public_release = [ False for i in range(num_images_per_camera_trajectory) ]
+        exclude_reason             = [ "OUTSIDE VIEWING AREA (BAD INITIALIZATION)" for i in range(num_images_per_camera_trajectory) ]
+    elif c["Scene type"] == "OUTSIDE VIEWING AREA (BAD TRAJECTORY)":
+        included_in_public_release = [ False for i in range(num_images_per_camera_trajectory) ]
+        exclude_reason             = [ "OUTSIDE VIEWING AREA (BAD TRAJECTORY)" for i in range(num_images_per_camera_trajectory) ]
+    else:
+        frames_ids_excluded_flagged_ = array(frames_ids_excluded_flagged[(scene_name, camera_name)])
+        included_in_public_release   = logical_not(in1d(frame_ids, frames_ids_excluded_flagged_))
+        exclude_reason               = [ "" if included_in_public_release[i] else "CONTENT FLAGGED FOR REMOVAL" for i in range(num_images_per_camera_trajectory) ]
+
+    df_ = pd.DataFrame(
+        columns=df_columns,
+        data={"scene_name"                 : scene_names,
+              "camera_name"                : camera_names,
+              "frame_id"                   : frame_ids,
+              "included_in_public_release" : included_in_public_release,
+              "exclude_reason"             : exclude_reason})
+
+    df = df.append(df_)
+
+included_in_public_release_counts = df.included_in_public_release.value_counts()
+print("[HYPERSIM: DATASET_GENERATE_IMAGE_METADATA] Images included in public release: " + str(included_in_public_release_counts[True]))
+
+
+
+df.to_csv(metadata_images_csv_file, index=False)
+
+
+
+print("[HYPERSIM: DATASET_GENERATE_IMAGE_METADATA] Finished.")

code/python/analysis/dataset_generate_scene_labeling_statistics.py

@@ -7,12 +7,9 @@
 
 import argparse
 import fnmatch
-import glob
-import h5py
 import inspect
 import os
 import pandas as pd
-import sklearn.preprocessing
 
 import path_utils
 

code/python/tools/dataset_copy_semantic_segmentations.py

@@ -51,14 +51,9 @@
     files_to_copy = [
         "mesh_objects_si.hdf5",
         "mesh_objects_sii.hdf5",
-        "metadata_groups.csv",
-        "metadata_materials.csv",
         "metadata_objects.csv",
         "metadata_scene_annotation_tool.log",
         "metadata_semantic_colors.hdf5",
-        "metadata_semantic_instance_bounding_box_object_aligned_2d_extents.hdf5",
-        "metadata_semantic_instance_bounding_box_object_aligned_2d_orientations.hdf5",
-        "metadata_semantic_instance_bounding_box_object_aligned_2d_positions.hdf5",
         "metadata_semantic_instance_colors.hdf5"
     ]