feat(fastalbel_to_t4dataset): fix keyframe handling

perception_dataset/deepen/deepen_to_t4_converter.py

@@ -21,7 +21,9 @@
 )
 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/fastlabel_to_t4/fastlabel_2d_to_t4_updater.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-import json
 import os.path as osp
 from pathlib import Path
 import shutil
@@ -40,10 +39,14 @@ def __init__(
 
     def convert(self) -> None:
         t4_datasets = sorted([d.name for d in self._input_base.iterdir() if d.is_dir()])
-        anno_jsons_dict = self._load_annotation_jsons(t4_datasets)
+        anno_jsons_dict = self._load_annotation_jsons(t4_datasets, "_CAM")
         fl_annotations = self._format_fastlabel_annotation(anno_jsons_dict)
 
         for t4dataset_name in t4_datasets:
+            # Check if annotation exists
+            if t4dataset_name not in fl_annotations.keys():
+                continue
+
             # Check if input directory exists
             input_dir = self._input_base / t4dataset_name
             input_annotation_dir = input_dir / "annotation"
@@ -78,28 +81,14 @@ def convert(self) -> None:
             else:
                 raise ValueError("If you want to overwrite files, use --overwrite option.")
 
-            if t4dataset_name not in fl_annotations.keys():
-                logger.warning(f"No annotation for {t4dataset_name}")
-                continue
-
             # Start updating annotations
-            annotation_files_updater = AnnotationFilesUpdater(description=self._description)
+            annotation_files_updater = AnnotationFilesUpdater(
+                description=self._description, surface_categories=self._surface_categories
+            )
             annotation_files_updater.convert_one_scene(
                 input_dir=input_dir,
                 output_dir=output_dir,
                 scene_anno_dict=fl_annotations[t4dataset_name],
                 dataset_name=t4dataset_name,
             )
-
-    def _load_annotation_jsons(self, t4_datasets: list[str]) -> dict[str, list[dict[str, any]]]:
-        anno_dict = {}
-        for file in self._input_anno_files:
-            t4_dataset_name = file.name.split("_CAM")[0]
-            if t4_dataset_name not in t4_datasets:
-                continue
-            with open(file) as f:
-                one_label = json.load(f)
-                if t4_dataset_name not in anno_dict.keys():
-                    anno_dict[t4_dataset_name] = []
-                anno_dict[t4_dataset_name].extend(one_label)
-        return anno_dict
+            logger.info(f"Finished updating annotations for {t4dataset_name}")

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"]
@@ -116,9 +117,7 @@ def convert_one_scene(
             prev_wid_hgt: Tuple = (0, 0)
             # NOTE: num_cameras is always 6, because it is hard coded above.
             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"]:
+                if sample_nuim["fileformat"] == "png" or sample_nuim["fileformat"] == "jpg":
                     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
 import os.path as osp
-from typing import Any, Dict
+from pathlib import Path
+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 +13,9 @@
     SurfaceAnnTable,
     VisibilityTable,
 )
+from perception_dataset.t4_dataset.resolver.keyframe_consistency_resolver import (
+    KeyFrameConsistencyResolver,
+)
 
 
 def _load_json(filepath: str) -> Any:
@@ -21,8 +25,13 @@ 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 +55,9 @@ 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),

perception_dataset/t4_dataset/classes/object_ann.py

@@ -96,7 +96,7 @@ def from_json(cls, filepath: str) -> ObjectAnnTable:
                 attribute_tokens=item["attribute_tokens"],
                 bbox=item["bbox"],
                 mask=item["mask"],
-                automatic_annotation=item["automatic_annotation"],
+                automatic_annotation=item.get("automatic_annotation", False),
             )
             record.token = item["token"]
             table.set_record_to_table(record)

perception_dataset/t4_dataset/classes/sample_annotation.py

@@ -201,7 +201,7 @@ def from_json(cls, filepath: str) -> SampleAnnotationTable:
                 },
                 num_lidar_pts=item["num_lidar_pts"],
                 num_radar_pts=item["num_radar_pts"],
-                automatic_annotation=item["automatic_annotation"],
+                automatic_annotation=item.get("automatic_annotation", False),
             )
             record.token = item["token"]
             table.set_record_to_table(record)

perception_dataset/t4_dataset/classes/surface_ann.py

@@ -67,7 +67,7 @@ def from_json(cls, filepath: str) -> SurfaceAnnTable:
                 category_token=item["category_token"],
                 mask=item["mask"],
                 sample_data_token=item["sample_data_token"],
-                automatic_annotation=item["automatic_annotation"],
+                automatic_annotation=item.get("automatic_annotation", False),
             )
             record.token = item["token"]
             table.set_record_to_table(record)