update new dataset

builder.py

@@ -3,83 +3,8 @@
 # depth:  2025_06_16_21_27-mastlab-yao-0001
 #
 import os
-from pandas.io.xml import file_exists
 import polars as pl
-from collections import defaultdict
-from polars.testing import assert_series_equal
-from datasets import (
-    DatasetBuilder,
-    BuilderConfig,
-    load_dataset_builder,
-    DatasetInfo,
-    Features,
-    Value,
-    SplitGenerator,
-    Split,
-)
-
-from datasets import DatasetBuilder, BuilderConfig, load_dataset_builder
-
-
-import polars as pl
-import datasets
-from datasets import GeneratorBasedBuilder, DatasetInfo, Features, Value, Split
-
-
-class PolarsBuilder(GeneratorBasedBuilder):
-    VERSION = datasets.Version("1.0.0")
-
-    @property
-    def df(self):
-        data_dir = self.config.data_dir
-        if data_dir is None:
-            raise ValueError("data_dir must be specified in the config")
-        return build_table(data_dir)
-
-    def _info(self):
-        feats = Features(
-            {
-                "time": Value(dtype="string", id=None),
-                "location_depth": Value(dtype="string", id=None),
-                "subject_depth": Value(dtype="string", id=None),
-                "index_rgb_depth": Value(dtype="int64", id=None),
-                "file_path_depth": Value(dtype="string", id=None),
-                "index_radar": Value(dtype="int64", id=None),
-                "location_rgb": Value(dtype="string", id=None),
-                "subject_rgb": Value(dtype="string", id=None),
-                "file_path_rgb": Value(dtype="string", id=None),
-                "location_radar": Value(dtype="string", id=None),
-                "subject_radar": Value(dtype="string", id=None),
-                "length_radar_cube": Value(dtype="int64", id=None),
-                "dimension_radar": Value(dtype="string", id=None),
-                "window_size_radar": Value(dtype="int64", id=None),
-                "window_stride_radar": Value(dtype="int64", id=None),
-                "selected_range_bin": Value(dtype="int64", id=None),
-                "enthrophy": Value(dtype="float32", id=None),
-                "file_path_radar": Value(dtype="string", id=None),
-            }
-        )
-
-        return DatasetInfo(
-            description="自定义 Builder，使用 Polars 自动推断 schema",
-            features=feats,
-            supervised_keys=None,
-        )
-
-    def _split_generators(self, dl_manager):
-        df = self.df
-        train = df
-        # train = df.sample(frac=0.9, with_replacement=False)
-        # val = df.filter(~pl.col("row_nr").is_in(train["row_nr"]))
-        return [
-            datasets.SplitGenerator(name=Split.TRAIN, gen_kwargs={"subset": train}),
-            # datasets.SplitGenerator(name=Split.VALIDATION, gen_kwargs={"subset": val}),
-        ]
-
-    def _generate_examples(self, subset: pl.DataFrame):
-        # 将 Polars DataFrame 转为 Python dict 行返回
-        for idx, row in enumerate(subset.to_dicts()):
-            yield idx, row
+import re
 
 
 def build_table(data_root):
@@ -87,42 +12,41 @@ def build_table(data_root):
     radar_table = build_single_table(data_root, "radardata", parse_radar_data)
     rgbdata_table = build_single_table(data_root, "rgbdata", parse_rgbdata_data)
 
-    depth_table = depth_table.with_columns(
-        ((pl.col("index_depth") + 1) // 2).alias("index_radar")
-    )
-    rgbdata_table = rgbdata_table.with_columns(
-        ((pl.col("index_rgb") + 1) // 2).alias("index_radar")
-    )
-    # print(radar_table)
-
+    print(len(depth_table), len(radar_table), len(rgbdata_table))
     t1 = depth_table.join(
         rgbdata_table,
-        left_on=["time", "index_depth", "index_radar"],
-        right_on=["time", "index_rgb", "index_radar"],
-    ).rename({"index_depth": "index_rgb_depth"})
+        left_on=["id", "index_depth"],
+        right_on=["id", "index_rgb"],
+    ).rename({"index_depth": "index_frame"})
 
-    # print(t1)
+    assert_columns_equal(t1, "time", "time_right")
+    t1 = t1.drop("time_right")
+
+    print(t1["index_frame"])
+
+    print(t1)
     # print(t1.columns)
 
     data_table = t1.join(
         radar_table,
-        left_on=["time", "index_radar"],
-        right_on=["time", "index_radar"],
+        left_on=["id", "index_frame"],
+        right_on=["id", "index_radar"],
         how="inner",
     )
-    data_table = data_table.rename(
-        {
-            # "file_path": "file_path_depthdata",
-            # "location": "location_depthdata",
-            # "subject": "subject_depthdata",
-            # "index_tmp": "index_rgb_depth",
-            # "index": "index_radar",
-        }
-    )
+
+    assert_columns_equal(data_table, "time", "time_right")
+    data_table = data_table.drop("time_right")
 
     return data_table
 
 
+def assert_columns_equal(df, col1, col2):
+    """Assert that two columns have identical values"""
+    are_equal = (df[col1] == df[col2]).all()
+    assert are_equal, f"Columns {col1} and {col2} are not identical"
+    print(f"✓ Columns {col1} and {col2} are identical")
+
+
 def build_single_table(data_root, source_relatvie_root, parse_func):
     data = []
     for file_name in os.listdir(os.path.join(data_root, source_relatvie_root)):
@@ -141,22 +65,25 @@ def build_single_table(data_root, source_relatvie_root, parse_func):
 
 def parse_depth_data(source_relatvie_root, file_name):
     """
-    2025_05_12_15_48-mastlab-yao-0001
+    769E33E0_2025_06_16_21_27-mastlab-yao-0001-1.npy
     """
     base_name = file_name.split(".")[0]
 
     # Parse the base name to extract the relevant information
     parts = base_name.split("-")
-    if len(parts) != 4:
+    if len(parts) != 5:
         raise ValueError(
             "Base name format is incorrect. Expected format: YYYY_MM_DD_HH_MM-<other_info>-<other_info>-<other_info>"
         )
 
+    id, time_part = split_id_time(parts[0])
     return dict(
-        time=parts[0],
+        id=id,
+        time=time_part,
         location_depth=parts[1],
         subject_depth=parts[2],
         index_depth=int(parts[3]),
+        sub_index_depth=int(parts[4]),
         file_path_depth=os.path.join(
             source_relatvie_root, file_name
         ),  # Use the relative path
@@ -165,19 +92,19 @@ def parse_depth_data(source_relatvie_root, file_name):
 
 def parse_radar_data(source_relatvie_root, file_name):
     """
-    2025_05_12_15_48-mastlab-yao-100-2D-2-2-0001
-    2025_06_16_21_27-mastlab-yao-1400-2D-2-2-0005-081-4.47
+    769E33E0_2025_06_16_21_27-mastlab-yao-1400-2D-2-2-0001-076-5.90.mat
     """
     base_name = file_name.split(".")[0]
     # Parse the base name to extract the relevant information
     parts = base_name.split("-")
     if len(parts) != 10:
         raise ValueError(
-            "Base name format is incorrect. Expected format: YYYY_MM_DD_HH_MM-<other_info>-<other_info>-<other_info>-<other_info>-<other_info>-<other_info>"
+            "Base name format is incorrect. Expected format: XXXXXXXX_YYYY_MM_DD_HH_MM-<other_info>-<other_info>-<other_info>-<other_info>-<other_info>-<other_info>"
         )
-
+    id, time_part = split_id_time(parts[0])
     return dict(
-        time=parts[0],
+        id=id,
+        time=time_part,
         location_radar=parts[1],
         subject_radar=parts[2],
         length_radar_cube=int(parts[3]),
@@ -196,31 +123,48 @@ def parse_radar_data(source_relatvie_root, file_name):
 
 def parse_rgbdata_data(source_relatvie_root, file_name):
     """
-    2025_05_12_15_48-mastlab-yao-0001
+    769E33E0_2025_06_16_21_27-mastlab-yao-0001-1.npy
     """
 
     base_name = file_name.split(".")[0]
 
     # Parse the base name to extract the relevant information
     parts = base_name.split("-")
-    if len(parts) != 4:
+    if len(parts) != 5:
         raise ValueError(
-            "Base name format is incorrect. Expected format: YYYY_MM_DD_HH_MM-<other_info>-<other_info>-<other_info>"
+            "Base name format is incorrect. Expected format: xxxx_YYYY_MM_DD_HH_MM-<other_info>-<other_info>-<other_info>-<other_info>"
         )
+    id, time_part = split_id_time(parts[0])
 
     return dict(
-        time=parts[0],
+        id=id,
+        time=time_part,
         location_rgb=parts[1],
         subject_rgb=parts[2],
         index_rgb=int(parts[3]),
+        sub_index_rgb=int(parts[4]),
         file_path_rgb=os.path.join(
             source_relatvie_root, file_name
         ),  # Use the relative path
     )
 
 
+def split_id_time(part):
+    pattern = r"^([0-9A-F]+)_(\d{4}_\d{2}_\d{2}_\d{2}_\d{2})$"
+    id_part, time_part = part.split("_", 1)
+    match = re.match(pattern, part)
+    if match:
+        id_part = match.group(1)
+        time_part = match.group(2)
+        return id_part, time_part
+    else:
+        raise ValueError("String format is incorrect")
+
+
 if __name__ == "__main__":
     data_root = os.path.dirname(os.path.abspath(__file__))
     data_table = build_table(data_root)
+    print(data_table)
+    print(data_table.columns)
     # data_table.write_parquet("./train.parquet")
     data_table.write_csv("./train.csv")