import os import random import datasets from datasets.tasks import ImageClassification _NAMES = { "PearlRiver": [2.33, 2.53, 2.37, 2.41], "YoungChang": [2.53, 2.63, 2.97, 2.71], "Steinway-T": [3.6, 3.63, 3.67, 3.63], "Hsinghai": [3.4, 3.27, 3.2, 3.29], "Kawai": [3.17, 2.5, 2.93, 2.87], "Steinway": [4.23, 3.67, 4, 3.97], "Kawai-G": [3.37, 2.97, 3.07, 3.14], "Yamaha": [3.23, 3.03, 3.17, 3.14], } _HOMEPAGE = f"https://www.modelscope.cn/datasets/ccmusic-database/{os.path.basename(__file__)[:-3]}" _DOMAIN = f"{_HOMEPAGE}/resolve/master/data" _PITCHES = { "009": "A2", "010": "A2#/B2b", "011": "B2", "100": "C1", "101": "C1#/D1b", "102": "D1", "103": "D1#/E1b", "104": "E1", "105": "F1", "106": "F1#/G1b", "107": "G1", "108": "G1#/A1b", "109": "A1", "110": "A1#/B1b", "111": "B1", "200": "C", "201": "C#/Db", "202": "D", "203": "D#/Eb", "204": "E", "205": "F", "206": "F#/Gb", "207": "G", "208": "G#/Ab", "209": "A", "210": "A#/Bb", "211": "B", "300": "c", "301": "c#/db", "302": "d", "303": "d#/eb", "304": "e", "305": "f", "306": "f#/gb", "307": "g", "308": "g#/ab", "309": "a", "310": "a#/bb", "311": "b", "400": "c1", "401": "c1#/d1b", "402": "d1", "403": "d1#/e1b", "404": "e1", "405": "f1", "406": "f1#/g1b", "407": "g1", "408": "g1#/a1b", "409": "a1", "410": "a1#/b1b", "411": "b1", "500": "c2", "501": "c2#/d2b", "502": "d2", "503": "d2#/e2b", "504": "e2", "505": "f2", "506": "f2#/g2b", "507": "g2", "508": "g2#/a2b", "509": "a2", "510": "a2#/b2b", "511": "b2", "600": "c3", "601": "c3#/d3b", "602": "d3", "603": "d3#/e3b", "604": "e3", "605": "f3", "606": "f3#/g3b", "607": "g3", "608": "g3#/a3b", "609": "a3", "610": "a3#/b3b", "611": "b3", "700": "c4", "701": "c4#/d4b", "702": "d4", "703": "d4#/e4b", "704": "e4", "705": "f4", "706": "f4#/g4b", "707": "g4", "708": "g4#/a4b", "709": "a4", "710": "a4#/b4b", "711": "b4", "800": "c5", } _URLS = { "audio": f"{_DOMAIN}/audio.zip", "mel": f"{_DOMAIN}/mel.zip", "eval": f"{_DOMAIN}/eval.zip", } class pianos(datasets.GeneratorBasedBuilder): def _info(self): names = list(_NAMES.keys()) if self.config.name == "default": names = names[:-1] return datasets.DatasetInfo( features=( datasets.Features( { "audio": datasets.Audio(sampling_rate=44100), "mel": datasets.Image(), "label": datasets.features.ClassLabel(names=names), "pitch": datasets.features.ClassLabel( names=list(_PITCHES.values()) ), "bass_score": datasets.Value("float32"), "mid_score": datasets.Value("float32"), "treble_score": datasets.Value("float32"), "avg_score": datasets.Value("float32"), } ) if self.config.name != "eval" else datasets.Features( { "mel": datasets.Image(), "label": datasets.features.ClassLabel(names=names), "pitch": datasets.features.ClassLabel( names=list(_PITCHES.values()) ), "bass_score": datasets.Value("float32"), "mid_score": datasets.Value("float32"), "treble_score": datasets.Value("float32"), "avg_score": datasets.Value("float32"), } ) ), homepage=_HOMEPAGE, license="CC-BY-NC-ND", version="1.2.0", supervised_keys=("mel", "label"), task_templates=ImageClassification( image_column="mel", label_column="label", ), ) def _split_generators(self, dl_manager): dataset = [] if self.config.name != "eval": subset = {} audio_files = dl_manager.download_and_extract(_URLS["audio"]) for path in dl_manager.iter_files([audio_files]): fname = os.path.basename(path) if fname.endswith(".wav"): lebal = os.path.basename(os.path.dirname(path)) if self.config.name == "default" and lebal == "Yamaha": continue subset[fname.split(".")[0]] = { "audio": path, "label": lebal, "pitch": _PITCHES[fname[1:4]], "bass_score": _NAMES[lebal][0], "mid_score": _NAMES[lebal][1], "treble_score": _NAMES[lebal][2], "avg_score": _NAMES[lebal][3], } mel_files = dl_manager.download_and_extract(_URLS["mel"]) for path in dl_manager.iter_files([mel_files]): fname = os.path.basename(path) pname = fname.split(".")[0] if fname.endswith(".jpg") and pname in subset: subset[pname]["mel"] = path dataset = list(subset.values()) else: data_files = dl_manager.download_and_extract(_URLS["eval"]) for path in dl_manager.iter_files([data_files]): fname: str = os.path.basename(path) if fname.endswith(".jpg"): lebal = os.path.basename(os.path.dirname(path)) dataset.append( { "mel": path, "label": lebal, "pitch": _PITCHES[fname.split("_")[0]], "bass_score": _NAMES[lebal][0], "mid_score": _NAMES[lebal][1], "treble_score": _NAMES[lebal][2], "avg_score": _NAMES[lebal][3], } ) names = list(_NAMES.keys()) if self.config.name == "default": names = names[:-1] categories = {} for name in names: categories[name] = [] for data in dataset: categories[data["label"]].append(data) testset, validset, trainset = [], [], [] for cls in categories: random.shuffle(categories[cls]) count = len(categories[cls]) p80 = int(count * 0.8) p90 = int(count * 0.9) trainset += categories[cls][:p80] validset += categories[cls][p80:p90] testset += categories[cls][p90:] 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, path in enumerate(files): yield i, path