Merge branch 'main' into feat/port/rosbag2_to_annotated_t4_tlr

docs/tracking_simulator_conversions.md

@@ -0,0 +1,25 @@
+# Tools Overview
+
+This document offers a concise overview of the scripts employed in the conversion process for the tracking simulator.  
+Prior to reading this document, we recommend referring to the [tools overview](./tools_overview.md) for a comprehensive description of the entire repository.
+
+## tracking debugger rosbag to annotated T4 format
+
+input: rosbag2 with ground_truth/objects  
+output: T4 format data
+
+```bash
+python -m perception_dataset.convert --config config/rosbag2_to_t4/convert_tracking_debugger_to_tracking_eval.yaml
+```
+
+## scenario simulator rosbag
+
+This guide emphasizes the utilization of the rosbag output generated by the planning scenario simulator.
+The T4 format of this output proves valuable for conducting regression tests on the tracking module.
+
+input: rosbag2 with ground_truth/objects and obstacle_segmentation/pointcloud  
+output: T4 format data
+
+```bash
+python -m perception_dataset.convert --config config/rosbag2_to_t4/convert_scenario_sim_to_tracking_eval.yaml
+```

perception_dataset/deepen/annotated_t4_to_deepen_converter.py

@@ -83,6 +83,11 @@ def _convert_one_scene(self, input_dir: str, scene_name: str):
 
                 sample_data_record = nusc.get("sample_data", sample_record["data"][sensor])
                 file_id = osp.basename(sample_data_record["filename"]).replace(".pcd.bin", ".pcd")
+
+                # Original T4 format names the file_id as 000000.pcd.bin for example.
+                # We need to convert it to 0.pcd in this case.
+                file_id = str(int(file_id.split(".")[0])) + ".pcd"
+
                 label_category_id = self._label_converter.convert_label(category_record["name"])
 
                 attributes_records = [

tests/config/convert_tracking_debugger_to_tracking_eval.yaml

@@ -0,0 +1,16 @@
+task: convert_rosbag2_to_t4_tracking
+description:
+  scene: ["tracking_regression", "synthetic"]
+conversion:
+  # path to rosbag dir output by simulator
+  input_base: ./data/rosbag2_tracking_debugger
+  output_base: ./data/pseudo_labeling_t4_format
+  skip_timestamp: 1.0
+  num_load_frames: 0
+  object_topic_name: /perception/object_recognition/ground_truth/objects
+  object_msg_type: TrackedObjects
+  lidar_sensor: # tracking debugger data has no lidar
+    topic: ""
+    channel: ""
+  camera_sensors: [] # synthetic data has no images
+  topic_list: ./config/topic_list_tracking_eval.yaml