sync

.gitattributes

@@ -1,35 +1,35 @@
 *.7z filter=lfs diff=lfs merge=lfs -text
 *.arrow filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
 *.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
 *.ftz filter=lfs diff=lfs merge=lfs -text
 *.gz filter=lfs diff=lfs merge=lfs -text
 *.h5 filter=lfs diff=lfs merge=lfs -text
 *.joblib filter=lfs diff=lfs merge=lfs -text
 *.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
 *.model filter=lfs diff=lfs merge=lfs -text
 *.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
 *.onnx filter=lfs diff=lfs merge=lfs -text
 *.ot filter=lfs diff=lfs merge=lfs -text
 *.parquet filter=lfs diff=lfs merge=lfs -text
 *.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
 *.pt filter=lfs diff=lfs merge=lfs -text
 *.pth filter=lfs diff=lfs merge=lfs -text
 *.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
 saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
 *.tflite filter=lfs diff=lfs merge=lfs -text
 *.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.AppImage filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,4 @@
+example/*
+*__pycache__*
+test.py
+rename.sh

README.md

@@ -1,10 +1,10 @@
 ---
 title: Piano Trans
-emoji: 📉
-colorFrom: yellow
-colorTo: purple
+emoji: 🎹🎵
+colorFrom: purple
+colorTo: gray
 sdk: gradio
-sdk_version: 4.36.1
+sdk_version: 4.24.0
 app_file: app.py
 pinned: false
 license: mit

app.py

@@ -0,0 +1,212 @@
+import os
+import re
+import json
+import torch
+import shutil
+import requests
+import gradio as gr
+from piano_transcription_inference import PianoTranscription, load_audio, sample_rate
+from modelscope import snapshot_download
+from tempfile import NamedTemporaryFile
+from pydub.utils import mediainfo
+from urllib.parse import urlparse
+from convert import midi2xml, xml2abc, xml2mxl, xml2jpg
+
+CACHE_DIR = "./flagged"
+WEIGHTS_PATH = (
+    snapshot_download("MuGeminorum/piano_transcription", cache_dir="./__pycache__")
+    + "/CRNN_note_F1=0.9677_pedal_F1=0.9186.pth"
+)
+
+
+def clean_cache(cache_dir=CACHE_DIR):
+    if os.path.exists(cache_dir):
+        shutil.rmtree(cache_dir)
+
+    os.mkdir(cache_dir)
+
+
+def get_audio_file_type(file_path: str):
+    try:
+        # 获取媒体信息
+        info = mediainfo(file_path)
+        # 返回文件格式
+        return "." + info["format_name"]
+
+    except Exception as e:
+        print(f"Error occurred: {e}")
+        return None
+
+
+def download_audio(url: str, save_path: str):
+    with NamedTemporaryFile(delete=False, suffix="_temp") as tmp_file:
+        temp_file_path = tmp_file.name
+        # 发送HTTP GET请求并下载内容
+        response = requests.get(url, stream=True)
+        # 检查请求是否成功
+        if response.status_code == 200:
+            # 将音频内容写入临时文件
+            for chunk in response.iter_content(chunk_size=8192):
+                tmp_file.write(chunk)
+
+        else:
+            print(f"Failed to download file: HTTP {response.status_code}")
+            return ""
+
+    ext = get_audio_file_type(temp_file_path)
+    full_path = f"{save_path}{ext}"
+    # 重命名临时文件以包含正确的扩展名
+    shutil.move(temp_file_path, full_path)
+    return full_path
+
+
+def is_url(s: str):
+    try:
+        # 解析字符串
+        result = urlparse(s)
+        # 检查scheme（如http, https）和netloc（域名）
+        return all([result.scheme, result.netloc])
+
+    except:
+        # 如果解析过程中发生异常，则返回False
+        return False
+
+
+def audio2midi(audio_path: str):
+    # Load audio
+    audio, _ = load_audio(audio_path, sr=sample_rate, mono=True)
+    # Transcriptor
+    transcriptor = PianoTranscription(
+        device="cuda" if torch.cuda.is_available() else "cpu",
+        checkpoint_path=WEIGHTS_PATH,
+    )
+    # device: 'cuda' | 'cpu' Transcribe and write out to MIDI file
+    midi_path = f"{CACHE_DIR}/output.mid"
+    # midi_path = audio_path.replace(audio_path.split(".")[-1], "mid")
+    transcriptor.transcribe(audio, midi_path)
+    return midi_path, os.path.basename(audio_path).split(".")[-2].capitalize()
+
+
+def upl_infer(audio_path: str):
+    clean_cache()
+    try:
+        midi, title = audio2midi(audio_path)
+        xml = midi2xml(midi, title)
+        abc = xml2abc(xml)
+        mxl = xml2mxl(xml)
+        pdf, jpg = xml2jpg(xml)
+        return midi, pdf, xml, mxl, abc, jpg
+
+    except Exception as e:
+        return None, None, None, None, f"{e}", None
+
+
+def get_first_integer(input_string: str):
+    match = re.search(r"\d+", input_string)
+    if match:
+        return str(int(match.group()))
+
+    else:
+        return ""
+
+
+def music163_song_info(id: str):
+    detail_api = "https://music.163.com/api/v3/song/detail"
+    parm_dict = {"id": id, "c": str([{"id": id}]), "csrf_token": ""}
+    free = False
+    song_name = "获取歌曲失败 Failed to get the song"
+    response = requests.get(detail_api, params=parm_dict)
+    # 检查请求是否成功
+    if response.status_code == 200:
+        # 处理成功响应
+        data = json.loads(response.text)
+        if data and "songs" in data and data["songs"]:
+            fee = int(data["songs"][0]["fee"])
+            free = fee == 0 or fee == 8
+            song_name = str(data["songs"][0]["name"])
+
+        else:
+            song_name = "歌曲不存在 Song not exist"
+
+    else:
+        raise ConnectionError(f"Error: {response.status_code}, {response.text}")
+
+    return song_name, free
+
+
+def url_infer(audio_url: str):
+    clean_cache()
+    song_name = ""
+    download_path = f"{CACHE_DIR}/output"
+    try:
+        if is_url(audio_url):
+            if "163" in audio_url and not audio_url.endswith(".mp3"):
+                song_id = get_first_integer(audio_url.split("?id=")[1])
+                audio_url = (
+                    f"https://music.163.com/song/media/outer/url?id={song_id}.mp3"
+                )
+                song_name, free = music163_song_info(song_id)
+                if not free:
+                    raise AttributeError("付费歌曲无法解析 Unable to parse VIP songs")
+
+            download_path = download_audio(audio_url, download_path)
+
+        midi, title = audio2midi(download_path)
+        if song_name:
+            title = song_name
+
+        xml = midi2xml(midi, title)
+        abc = xml2abc(xml)
+        mxl = xml2mxl(xml)
+        pdf, jpg = xml2jpg(xml)
+        return download_path, midi, pdf, xml, mxl, abc, jpg
+
+    except Exception as e:
+        return None, None, None, None, None, f"{e}", None
+
+
+if __name__ == "__main__":
+    with gr.Blocks() as iface:
+        with gr.Tab("上传模式 (Upload Mode)"):
+            gr.Interface(
+                fn=upl_infer,
+                inputs=gr.Audio(
+                    label="上传音频 (Upload an audio)",
+                    type="filepath",
+                ),
+                outputs=[
+                    gr.File(label="下载 MIDI (Download MIDI)"),
+                    gr.File(label="下载 PDF 乐谱 (Download PDF score)"),
+                    gr.File(label="下载 MusicXML (Download MusicXML)"),
+                    gr.File(label="下载 MXL (Download MXL)"),
+                    gr.Textbox(label="abc 乐谱 (abc notation)", show_copy_button=True),
+                    gr.Image(label="五线谱 (Staff)", type="filepath"),
+                ],
+                title="请上传音频 100% 后再点提交<br>Please make sure the audio is completely uploaded before clicking Submit",
+                allow_flagging="never",
+            )
+
+        with gr.Tab("直链模式 (Direct Link Mode)"):
+            gr.Interface(
+                fn=url_infer,
+                inputs=gr.Textbox(label="输入音频直链 URL (Input audio direct link)"),
+                outputs=[
+                    gr.Audio(label="下载音频 (Download audio)", type="filepath"),
+                    gr.File(label="下载 MIDI (Download MIDI)"),
+                    gr.File(label="下载 PDF 乐谱 (Download PDF score)"),
+                    gr.File(label="下载 MusicXML (Download MusicXML)"),
+                    gr.File(label="下载 MXL (Download MXL)"),
+                    gr.Textbox(label="abc 乐谱 (abc notation)", show_copy_button=True),
+                    gr.Image(label="五线谱 (Staff)", type="filepath"),
+                ],
+                title="网易云音乐可直接输入非 VIP 歌曲页面链接自动解析<br>For Netease Cloud music, you can directly input the non-VIP song page link",
+                examples=[
+                    "https://music.163.com/#/song?id=1945798894",
+                    "https://music.163.com/#/song?id=1945798973",
+                    "https://music.163.com/#/song?id=1946098771",
+                ],
+                allow_flagging="never",
+                cache_examples=False,
+            )
+
+    iface.launch()

convert.py

@@ -0,0 +1,132 @@
+import os
+import sys
+import fitz
+import requests
+import subprocess
+from PIL import Image
+from music21 import converter
+
+
+def download(url: str, directory: str, filename: str):
+    if directory != "" and not os.path.exists(directory):
+        os.makedirs(directory)
+    # Create the full path for the file to be saved
+    file_path = os.path.join(directory, filename)
+    # Send a GET request to the URL
+    response = requests.get(url, stream=True)
+    # Check if the request was successful
+    if response.status_code == 200:
+        # Open the file in write-binary mode
+        with open(file_path, "wb") as file:
+            # Write the contents of the response to the file
+            for chunk in response.iter_content(chunk_size=1024):
+                if chunk:  # Filter out keep-alive new chunks
+                    file.write(chunk)
+
+        print(f"The file has been downloaded and saved to {file_path}")
+
+    else:
+        print(f"Failed to download the file. Status code: {response.status_code}")
+
+    return os.path.join(directory, filename)
+
+
+if sys.platform.startswith("linux"):
+    apkname = "MuseScore.AppImage"
+    extra_dir = "squashfs-root"
+    if not os.path.exists(apkname):
+        download(
+            url="https://master.dl.sourceforge.net/project/musescore.mirror/v4.2.0/MuseScore-4.2.0.233521125-x86_64.AppImage?viasf=1",
+            directory="./",
+            filename=apkname,
+        )
+
+    if not os.path.exists(extra_dir):
+        subprocess.run(["chmod", "+x", f"./{apkname}"])
+        subprocess.run([f"./{apkname}", "--appimage-extract"])
+
+    MSCORE = f"./{extra_dir}/AppRun"
+    os.environ["QT_QPA_PLATFORM"] = "offscreen"
+
+else:
+    MSCORE = "D:/Program Files/MuseScore 3/bin/MuseScore3.exe"
+
+
+def add_title_to_xml(xml_path: str, title: str):
+    midi_data = converter.parse(xml_path)
+    # 将标题添加到 MIDI 文件中
+    midi_data.metadata.movementName = title
+    midi_data.metadata.composer = "Transcripted by AI"
+    # 保存修改后的 MIDI 文件
+    midi_data.write("musicxml", fp=xml_path)
+
+
+def xml2abc(xml_path: str):
+    result = subprocess.run(
+        ["python", "xml2abc.py", xml_path], stdout=subprocess.PIPE, text=True
+    )
+    if result.returncode == 0:
+        return result.stdout
+
+    return ""
+
+
+def xml2mxl(xml_path: str):
+    mxl_file = xml_path.replace(".musicxml", ".mxl")
+    command = [MSCORE, "-o", mxl_file, xml_path]
+    result = subprocess.run(command)
+    print(result)
+    return mxl_file
+
+
+def midi2xml(mid_file: str, title: str):
+    xml_file = mid_file.replace(".mid", ".musicxml")
+    command = [MSCORE, "-o", xml_file, mid_file]
+    result = subprocess.run(command)
+    add_title_to_xml(xml_file, title)
+    print(result)
+    return xml_file
+
+
+def xml2midi(xml_file: str):
+    midi_file = xml_file.replace(".musicxml", ".mid")
+    command = [MSCORE, "-o", midi_file, xml_file]
+    result = subprocess.run(command)
+    print(result)
+    return midi_file
+
+
+def pdf2img(pdf_path: str):
+    output_path = pdf_path.replace(".pdf", ".jpg")
+    doc = fitz.open(pdf_path)
+    # 创建一个图像列表
+    images = []
+    for page_number in range(doc.page_count):
+        page = doc[page_number]
+        # 将页面渲染为图像
+        image = page.get_pixmap()
+        # 将图像添加到列表
+        images.append(
+            Image.frombytes("RGB", [image.width, image.height], image.samples)
+        )
+    # 竖向合并图像
+    merged_image = Image.new(
+        "RGB", (images[0].width, sum(image.height for image in images))
+    )
+    y_offset = 0
+    for image in images:
+        merged_image.paste(image, (0, y_offset))
+        y_offset += image.height
+    # 保存合并后的图像为JPG
+    merged_image.save(output_path, "JPEG")
+    # 关闭PDF文档
+    doc.close()
+    return output_path
+
+
+def xml2jpg(xml_file: str):
+    pdf_score = xml_file.replace(".musicxml", ".pdf")
+    command = [MSCORE, "-o", pdf_score, xml_file]
+    result = subprocess.run(command)
+    print(result)
+    return pdf_score, pdf2img(pdf_score)

requirements.txt

@@ -0,0 +1,7 @@
+librosa==0.9.2
+piano_transcription_inference
+pymupdf
+music21
+modelscope
+torch
+transformers