feat(deepen_to_t4): add paint label downloader

config/convert_deepen_to_t4_paint_sample.yaml

@@ -0,0 +1,19 @@
+task: convert_deepen_to_t4
+description:
+  visibility:
+    full: "No occlusion of the object."
+    most: "Object is occluded, but by less than 50%."
+    partial: "The object is occluded by more than 50% (but not completely)."
+    none: "The object is 90-100% occluded and no points/pixels are visible in the label."
+  camera_index:
+    CAM_FRONT: 0
+    CAM_FRONT_RIGHT: 1
+    CAM_BACK_RIGHT: 2
+    CAM_BACK: 3
+    CAM_BACK_LEFT: 4
+    CAM_FRONT_LEFT: 5
+
+conversion:
+  input_base: ./data/non_annotated_t4_format
+  dataset_corresponding:
+    Dataset_name: dataset_id_in_Deepen_AI

perception_dataset/deepen/download_paint_annotations_to_pcd.py

@@ -0,0 +1,106 @@
+import argparse
+from datetime import date
+import json
+import os
+import zlib
+
+import numpy as np
+import requests
+import tqdm
+import yaml
+
+if os.environ.get("DEEPEN_CLIENT_ID") is None or os.environ.get("DEEPEN_ACCESS_TOKEN") is None:
+    raise ValueError(
+        'You need to properly set the environment variables "DEEPEN_CLIENT_ID" and "DEEPEN_ACCESS_TOKEN"'
+    )
+else:
+    CLIENT_ID = os.environ["DEEPEN_CLIENT_ID"]
+    ACCESS_TOKEN = os.environ["DEEPEN_ACCESS_TOKEN"]
+
+today = str(date.today()).replace("-", "")
+NUM_DIMENSIONS = 5
+
+
+def get_datasets(dataset_id: str, input_base_dir: str) -> None:
+    headers = {
+        "Authorization": f"Bearer {ACCESS_TOKEN}",
+        "Content-Type": "application/json",
+    }
+
+    if dataset_id is None:
+        print("Dataset ID not found")
+        return
+
+    DATASET_URL = f"https://tools.deepen.ai/api/v2/datasets/{dataset_id}/label_types/3d_point/paint_labels?stageId=QA"
+
+    try:
+        response = requests.get(DATASET_URL, headers=headers)
+        response.raise_for_status()
+    except requests.exceptions.RequestException as e:
+        raise SystemExit(e)
+
+    decompress_data = bytearray(zlib.decompress(bytearray(response.content)))
+
+    header_list = list(response.headers.values())
+    label_info = json.loads(header_list[3])
+    frame_size = list(label_info["frame_sizes"])
+
+    # [
+    #     [l,l,l,l,...],
+    #     [l,l,l,l,...],
+    #     ...
+    # ]
+
+    label_list = []
+    offset = 0
+    for i in range(len(frame_size)):
+        offset += frame_size[i - 1] if i != 0 else 0
+        label_list.append(
+            np.array([decompress_data[offset + j] for j in range(frame_size[i])], dtype=np.uint32)
+        )
+
+    sample_data_file = os.path.join(input_base_dir, "annotation", "sample_data.json")
+    sample_data = json.load(open(sample_data_file, "r"))
+    sample_data = list(filter(lambda d: d["filename"].split(".")[-2] == "pcd", sample_data))
+
+    for i in tqdm.tqdm(range(0, len(sample_data))):
+        pcd_file_path = os.path.join(input_base_dir, sample_data[i]["filename"])
+        scan = np.fromfile(pcd_file_path, dtype=np.float32)
+        points: np.ndarray = scan.reshape((-1, NUM_DIMENSIONS))  # 行数，列数
+        labelled_points: np.ndarray = np.concatenate(
+            [points, label_list[i].reshape(-1, 1)], 1, dtype=np.float32
+        )
+        labelled_points.tofile(pcd_file_path)
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--config",
+        type=str,
+        default="config/convert_deepen_to_t4.yaml",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default=".",
+        help="the directory where the annotation file is saved.",
+    )
+    args = parser.parse_args()
+
+    with open(args.config) as f:
+        config = yaml.safe_load(f)
+
+    assert (
+        config["task"] == "convert_deepen_to_t4"
+    ), f"use config file of convert_deepen_to_t4 task: {config['task']}"
+    dataset_id = list(config["conversion"]["dataset_corresponding"].values())
+    input_base_dir = os.path.join(
+        config["conversion"]["input_base"], os.listdir(config["conversion"]["input_base"])[0]
+    )
+
+    get_datasets(dataset_id[0], input_base_dir)
+
+
+if __name__ == "__main__":
+    main()