syncs

CTIS.py

@@ -2,7 +2,7 @@ import os
 import random
 import hashlib
 import datasets
-from datasets.tasks import AudioClassification
+from datasets.tasks import ImageClassification
 
 
 _HOMEPAGE = f"https://www.modelscope.cn/datasets/ccmusic-database/{os.path.basename(__file__)[:-3]}"
@@ -98,9 +98,9 @@ _NAMES = {
     "D0176": ["福(新)", "fu2_(reformed)"],
     "D0177": ["禄(新)", "lu4_(reformed)"],
     "D0178": ["寿(新)", "shou4_(reformed)"],
-    "D0179": ["宜春三星鼓福鼓老鼓", "yi2_chun1_san1_xing1_gu3_fu2_gu3_lao3_gu3"],
-    "D0180": ["宜春三星鼓禄鼓老鼓", "yi2_chun1_san1_xing1_gu3_lu4_gu3_lao3_gu3"],
-    "D0181": ["宜春三星鼓寿鼓老鼓", "yi2_chun1_san1_xing1_gu3_shou4_gu3_lao3_gu3"],
+    "D0179": ["宜春三星鼓福鼓老鼓", "yi2_chun1_san1_xing1_gu3_fu2_gu3_(traditional)"],
+    "D0180": ["宜春三星鼓禄鼓老鼓", "yi2_chun1_san1_xing1_gu3_lu4_gu3_(traditional)"],
+    "D0181": ["宜春三星鼓寿鼓老鼓", "yi2_chun1_san1_xing1_gu3_shou4_gu3_(traditional)"],
     "D0184": ["宜春三星鼓双铛", "yi2_chun1_san1_xing1_gu3_shuang1_ding1"],
     "D0185": ["宜春三星鼓单铛", "yi2_chun1_san1_xing1_gu3_dan1_ding1"],
     "D0186": ["宜春三星鼓镲", "yi2_chun1_san1_xing1_gu3_chao3"],
@@ -234,29 +234,47 @@ _NAMES = {
 _URLS = {
     "audio": f"{_DOMAIN}/audio.zip",
     "mel": f"{_DOMAIN}/mel.zip",
+    "eval": f"{_DOMAIN}/eval.zip",
 }
 
 
 class CTIS(datasets.GeneratorBasedBuilder):
     def _info(self):
         return datasets.DatasetInfo(
-            features=datasets.Features(
-                {
-                    "audio": datasets.Audio(sampling_rate=44100),
-                    "mel": datasets.Image(),
-                    "label": datasets.features.ClassLabel(names=list(_NAMES.keys())),
-                    "cname": datasets.Value("string"),
-                    "pinyin": datasets.Value("string"),
-                }
+            features=(
+                datasets.Features(
+                    {
+                        "audio": datasets.Audio(sampling_rate=44100),
+                        "mel": datasets.Image(),
+                        "label": datasets.features.ClassLabel(
+                            names=list(_NAMES.keys())
+                        ),
+                        "cname": datasets.Value("string"),
+                        "pinyin": datasets.Value("string"),
+                    }
+                )
+                if self.config.name == "default"
+                else (
+                    datasets.Features(
+                        {
+                            "mel": datasets.Image(),
+                            "cqt": datasets.Image(),
+                            "chroma": datasets.Image(),
+                            "label": datasets.features.ClassLabel(
+                                names=list(_NAMES.keys())
+                            ),
+                        }
+                    )
+                )
             ),
-            supervised_keys=("audio", "label"),
+            supervised_keys=("mel", "label"),
             homepage=_HOMEPAGE,
             license="CC-BY-NC-ND",
             version="1.2.0",
             task_templates=[
-                AudioClassification(
-                    task="audio-classification",
-                    audio_column="audio",
+                ImageClassification(
+                    task="image-classification",
+                    image_column="mel",
                     label_column="label",
                 )
             ],
@@ -268,40 +286,90 @@ class CTIS(datasets.GeneratorBasedBuilder):
         return md5_obj.hexdigest()
 
     def _split_generators(self, dl_manager):
-        audio_files = dl_manager.download_and_extract(_URLS["audio"])
-        mel_files = dl_manager.download_and_extract(_URLS["mel"])
-        files = {}
-        for fpath in dl_manager.iter_files([audio_files]):
-            fname: str = os.path.basename(fpath)
-            label = os.path.basename(os.path.dirname(fpath))
-            if fname.endswith(".wav") and label in _NAMES and not "采访" in fname:
-                cls = os.path.basename(os.path.dirname(fpath)) + "/"
-                item_id = self._str2md5(cls + fname.split(".wa")[0])
-                files[item_id] = {"audio": fpath, "label": label}
+        if self.config.name == "default":
+            files = {}
+            audio_files = dl_manager.download_and_extract(_URLS["audio"])
+            mel_files = dl_manager.download_and_extract(_URLS["mel"])
+            for fpath in dl_manager.iter_files([audio_files]):
+                fname = os.path.basename(fpath)
+                if fname.endswith(".wav"):
+                    label = os.path.basename(os.path.dirname(fpath))
+                    item_id = self._str2md5(label + fname.split(".wa")[0])
+                    files[item_id] = {"audio": fpath}
 
-        for fpath in dl_manager.iter_files([mel_files]):
-            fname = os.path.basename(fpath)
-            label = os.path.basename(os.path.dirname(fpath))
-            if fname.endswith(".jpg") and label in _NAMES:
-                cls = os.path.basename(os.path.dirname(fpath)) + "/"
-                item_id = self._str2md5(cls + fname.split(".jp")[0])
-                files[item_id]["mel"] = fpath
+            for fpath in dl_manager.iter_files([mel_files]):
+                fname = os.path.basename(fpath)
+                if fname.endswith(".jpg"):
+                    label = os.path.basename(os.path.dirname(fpath))
+                    item_id = self._str2md5(label + fname.split(".jp")[0])
+                    files[item_id]["mel"] = fpath
 
-        dataset = list(files.values())
-        random.shuffle(dataset)
-        return [
-            datasets.SplitGenerator(
-                name=datasets.Split.TRAIN,
-                gen_kwargs={"files": dataset},
-            ),
-        ]
+            dataset = list(files.values())
+            random.shuffle(dataset)
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={"files": dataset},
+                ),
+            ]
+
+        else:
+            files = {key: [] for key in _NAMES}
+            data_files = dl_manager.download_and_extract(_URLS["eval"])
+            for fpath in dl_manager.iter_files([data_files]):
+                fname: str = os.path.basename(fpath)
+                if fname.endswith(".jpg") and "mel" in fpath:
+                    label = os.path.basename(os.path.dirname(fpath))
+                    files[label].append(fpath)
+
+            trainset, validset, testset = [], [], []
+            for cls in files:
+                random.shuffle(files[cls])
+                count = len(files[cls])
+                if count < 3:
+                    raise ValueError(f"Class {cls} has items < 3 !")
+
+                p10 = max(count // 10, 1)
+                p20 = p10 * 2
+                trainset += files[cls][p20:]
+                validset += files[cls][p10:p20]
+                testset += files[cls][:p10]
+
+            random.shuffle(trainset)
+            random.shuffle(validset)
+            random.shuffle(testset)
+            return [
+                datasets.SplitGenerator(
+                    name=datasets.Split.TRAIN,
+                    gen_kwargs={"files": trainset},
+                ),
+                datasets.SplitGenerator(
+                    name=datasets.Split.VALIDATION,
+                    gen_kwargs={"files": validset},
+                ),
+                datasets.SplitGenerator(
+                    name=datasets.Split.TEST,
+                    gen_kwargs={"files": testset},
+                ),
+            ]
 
     def _generate_examples(self, files):
-        for i, item in enumerate(files):
-            yield i, {
-                "audio": item["audio"],
-                "mel": item["mel"],
-                "label": item["label"],
-                "cname": _NAMES[item["label"]][0],
-                "pinyin": _NAMES[item["label"]][1],
-            }
+        if self.config.name == "default":
+            for i, item in enumerate(files):
+                label = os.path.basename(os.path.dirname(item["audio"]))
+                yield i, {
+                    "audio": item["audio"],
+                    "mel": item["mel"],
+                    "label": label,
+                    "cname": _NAMES[label][0],
+                    "pinyin": _NAMES[label][1],
+                }
+
+        else:
+            for i, item in enumerate(files):
+                yield i, {
+                    "mel": item,
+                    "cqt": item.replace("mel", "cqt"),
+                    "chroma": item.replace("mel", "chroma"),
+                    "label": os.path.basename(os.path.dirname(item)),
+                }