change to non-keyframe for no-annotation frame

Signed-off-by: Shunsuke Miura <[email protected]>

perception_dataset/deepen/deepen_to_t4_converter.py

@@ -21,7 +21,7 @@
 )
 from perception_dataset.rosbag2.rosbag2_converter import Rosbag2Converter
 from perception_dataset.t4_dataset.annotation_files_generator import AnnotationFilesGenerator
-from perception_dataset.t4_dataset.keyframe_consistency_resolver import KeyFrameConsistencyResolver
+from perception_dataset.t4_dataset.resolver.keyframe_consistency_resolver import KeyFrameConsistencyResolver
 from perception_dataset.utils.logger import configure_logger
 import perception_dataset.utils.misc as misc_utils
 

perception_dataset/t4_dataset/annotation_files_generator.py

@@ -87,10 +87,11 @@ def convert_one_scene(
 
         nusc = NuScenes(version="annotation", dataroot=input_dir, verbose=False)
         frame_index_to_sample_token: Dict[int, str] = {}
-        for frame_index, sample in enumerate(nusc.sample):
-            frame_index_to_sample_token[frame_index] = sample["token"]
+        for sample_data in nusc.sample_data:
+            frame_index = int((sample_data["filename"].split("/")[2]).split(".")[0])
+            frame_index_to_sample_token[frame_index] = sample_data["sample_token"]
         try:
-            if "LIDAR_TOP" in sample["data"]:
+            if "LIDAR_TOP" in nusc.sample[0]["data"]:
                 lidar_sensor_channel = SENSOR_ENUM.LIDAR_TOP.value["channel"]
             else:
                 lidar_sensor_channel = SENSOR_ENUM.LIDAR_CONCAT.value["channel"]
@@ -118,7 +119,7 @@ def convert_one_scene(
             for frame_index_nuim, sample_nuim in enumerate(nuim.sample_data):
                 if (
                     sample_nuim["fileformat"] == "png" or sample_nuim["fileformat"] == "jpg"
-                ) and sample_nuim["is_key_frame"]:
+                ):
                     cam = sample_nuim["filename"].split("/")[1]
                     cam_idx = self._camera2idx[cam]
 

perception_dataset/t4_dataset/annotation_files_updater.py

@@ -1,6 +1,7 @@
 import json
+from pathlib import Path
 import os.path as osp
-from typing import Any, Dict
+from typing import Any, Dict, List
 
 from perception_dataset.t4_dataset.annotation_files_generator import AnnotationFilesGenerator
 from perception_dataset.t4_dataset.classes import (
@@ -12,6 +13,7 @@
     SurfaceAnnTable,
     VisibilityTable,
 )
+from perception_dataset.t4_dataset.resolver.keyframe_consistency_resolver import KeyFrameConsistencyResolver
 
 
 def _load_json(filepath: str) -> Any:
@@ -21,8 +23,8 @@ def _load_json(filepath: str) -> Any:
 
 
 class AnnotationFilesUpdater(AnnotationFilesGenerator):
-    def __init__(self, with_camera: bool = True, description: Dict[str, Dict[str, str]] = ...):
-        super().__init__(with_camera, description)
+    def __init__(self, with_camera: bool = True, description: Dict[str, Dict[str, str]] = ..., surface_categories: List[str] = []):
+        super().__init__(with_camera, description, surface_categories)
         self.description = description
 
     def convert_one_scene(
@@ -46,6 +48,9 @@ def convert_one_scene(
             dataset_name=dataset_name,
         )
 
+        modifier = KeyFrameConsistencyResolver()
+        modifier.inspect_and_fix_t4_segment(Path(output_dir))
+
     def _init_table_from_json(self, anno_dir: str) -> None:
         self._attribute_table = AttributeTable.from_json(
             filepath=osp.join(anno_dir, AttributeTable.FILENAME),