upl piano trans

.gitattributes

@@ -1,37 +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
-*.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
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack 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
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* 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
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip 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
-*.gguf* filter=lfs diff=lfs merge=lfs -text
-*.ggml filter=lfs diff=lfs merge=lfs -text
-*.llamafile* filter=lfs diff=lfs merge=lfs -text
+*.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
+*.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
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack 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
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* 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
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip 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

@@ -1,2 +1,5 @@
-test.py
-rename.sh
+example/*
+*__pycache__*
+test.py
+rename.sh
+flagged/*

README.md

@@ -1,14 +1,12 @@
 ---
-title: Keep Spaces Active
-emoji: 📊
-colorFrom: gray
+title: Piano transcription
+emoji: 🎹🎵
+colorFrom: purple
 colorTo: gray
 sdk: gradio
 sdk_version: 5.9.1
 app_file: app.py
 pinned: false
-license: apache-2.0
-short_description: Keep your spaces active without payment
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+license: mit
+short_description: Piano Transcription Tool
+---

app.py

@@ -1,293 +1,215 @@
 import os
 import re
 import json
-import time
-import random
-import string
+import torch
+import shutil
 import requests
-import schedule
 import gradio as gr
-import pandas as pd
-from tqdm import tqdm
-from functools import partial
-from datetime import datetime, timedelta
-
-TIMEOUT = 15
-DELAY = 1
-
-
-def start_monitor(url: str):
-    payload = {
-        "data": ["", ""],
-        "event_data": None,  # 使用None来表示null
-        "fn_index": 0,
-        "trigger_id": 11,
-        "session_hash": "".join(
-            random.choice(string.ascii_lowercase) for _ in range(11)
-        ),
-    }
-    response = requests.post(f"{url}/queue/join?", json=payload)
-    # 检查请求是否成功
-    if response.status_code == 200:
-        return "monitoring"
+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(
+        "Genius-Society/piano_transcription",
+        cache_dir="./__pycache__",
+    )
+    + "/CRNN_note_F1=0.9677_pedal_F1=0.9186.pth"
+)
+
 
-    return "running"
+def clean_cache(cache_dir=CACHE_DIR):
+    if os.path.exists(cache_dir):
+        shutil.rmtree(cache_dir)
 
+    os.mkdir(cache_dir)
 
-def add_six_hours(match):
-    datetime_str = match.group(0)
-    dt = datetime.strptime(datetime_str, "%Y-%m-%d %H:%M:%S")
-    dt_plus_six = dt + timedelta(hours=6)
-    return dt_plus_six.strftime("%Y-%m-%d %H:%M:%S")
 
+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"]
 
-def fix_datetime(text: str):
-    datetime_pattern = r"\b\d{4}-\d{2}-\d{2} \d{2}:\d{2}:\d{2}\b"
-    return re.sub(datetime_pattern, add_six_hours, text)
+    except Exception as e:
+        print(f"Error occurred: {e}")
+        return None
 
 
-def get_studios(username: str):
-    # 请求负载
-    payload = {
-        "PageNumber": 1,
-        "PageSize": 1000,
-        "Name": "",
-        "SortBy": "gmt_modified",
-        "Order": "desc",
-    }
-    try:
-        # 发送PUT请求
-        response = requests.put(
-            f"https://www.modelscope.cn/api/v1/studios/{username}/list",
-            data=json.dumps(payload),
-            headers={
-                "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.3"
-            },
-            timeout=TIMEOUT,
-        )
+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
-        response.raise_for_status()
-        # 解析JSON响应
-        spaces: list = response.json()["Data"]["Studios"]
-        if spaces:
-            studios = []
-            for space in spaces:
-                studios.append(
-                    f"https://www.modelscope.cn/api/v1/studio/{username}/{space['Name']}/start_expired"
-                )
+        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)
 
-            return studios
+        else:
+            print(f"Failed to download file: HTTP {response.status_code}")
+            return ""
 
-    except requests.exceptions.Timeout as errt:
-        print(f"请求超时: {errt}, retrying...")
-        time.sleep(DELAY)
-        return get_studios(username)
+    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
 
-    except Exception as err:
-        print(f"请求发生错误: {err}")
 
-    return []
+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 get_spaces(username: str):
+def upl_infer(audio_path: str):
+    clean_cache()
     try:
-        # 发送GET请求
-        response = requests.get(
-            "https://huggingface.co/spaces-json",
-            params={"sort": "trending", "search": username},
-            headers={
-                "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537"
-            },
-            timeout=TIMEOUT,
-        )
-        # 检查请求是否成功
-        response.raise_for_status()
-        # 解析JSON响应
-        spaces: list = response.json()["spaces"]
-        studios = []
-        for space in spaces:
-            if space["author"] == username:
-                studios.append(
-                    f"https://{space['id'].replace('/', '-').replace('_', '-')}.hf.space"
-                )
+        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
 
-        return studios
+    except Exception as e:
+        return None, None, None, None, f"{e}", None
 
-    except requests.exceptions.Timeout as errt:
-        print(f"请求超时: {errt}, retrying...")
-        time.sleep(DELAY)
-        return get_spaces(username)
 
-    except Exception as err:
-        print(f"请求发生错误: {err}")
+def get_first_integer(input_string: str):
+    match = re.search(r"\d+", input_string)
+    if match:
+        return str(int(match.group()))
 
-    return []
+    else:
+        return ""
 
 
-def activate_space(url: str):
-    status = "running"
-    try:
-        if ".hf.space" in url:
-            response = requests.get(url, timeout=TIMEOUT)
-            response.raise_for_status()
-            if "-keep-spaces-active.hf.space" in url:
-                status = start_monitor(url)
+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:
-            response = requests.put(url, timeout=TIMEOUT)
-            response.raise_for_status()
-            print("Expired studio found, restarting...")
-            while (
-                requests.get(
-                    url.replace("/start_expired", "/status"),
-                    timeout=TIMEOUT,
-                ).json()["Data"]["Status"]
-                == "ExpiredCreating"
-            ):
-                requests.get(
-                    url.replace("/api/v1/studio/", "/studios/").replace(
-                        "/start_expired", ""
-                    ),
-                    timeout=TIMEOUT,
-                )
-                time.sleep(5)
+            song_name = "歌曲不存在 Song not exist"
 
-    except requests.exceptions.Timeout as e:
-        if ".hf.space" in url:
-            status = "restarting"
-        else:
-            print(f"Failed to activate {url} : {e}, retrying...")
-            return activate_space(url)
-
-    except requests.RequestException as e:
-        if (
-            "500 Server Error:" in f"{e}"
-            and response.json()["Message"] == "studio is not expired"
-        ):
-            status = "running"
-        else:
-            status = f"{e}"
+    else:
+        raise ConnectionError(f"Error: {response.status_code}, {response.text}")
 
-    except Exception as e:
-        status = f"{e}"
+    return song_name, free
 
-    return status
 
+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")
 
-def activate(hf_users: str, ms_users: str):
-    if not hf_users:
-        hf_users = os.getenv("hf_users")
+            download_path = download_audio(audio_url, download_path)
 
-    if not ms_users:
-        ms_users = hf_users
+        midi, title = audio2midi(download_path)
+        if song_name:
+            title = song_name
 
-    hf_usernames = hf_users.split(";")
-    ms_usernames = ms_users.split(";")
-    spaces = []
-    for user in tqdm(hf_usernames, desc="Collecting spaces..."):
-        username = user.strip()
-        if username:
-            spaces += get_spaces(username)
-            time.sleep(DELAY)
+        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
 
-    monitors, studios = [], []
-    for space in spaces:
-        if "keep-spaces-active" in space:
-            monitors.append(space)
-        else:
-            studios.append(space)
-    spaces = monitors + studios
-
-    for user in tqdm(ms_usernames, desc="Collecting studios..."):
-        username = user.strip()
-        if username:
-            spaces += get_studios(username)
-            time.sleep(DELAY)
-
-    output = []
-    for space in tqdm(spaces, desc="Activating spaces..."):
-        output.append(
-            {
-                "space": space.split("//")[-1].replace(
-                    "www.modelscope.cn/api/v1/studio/", ""
-                ),
-                "status": activate_space(space),
-            }
-        )
-        time.sleep(DELAY)
-
-    print("Activation complete!")
-    return pd.DataFrame(output)
-
-
-def monitor(hf_users: str, ms_users: str, period=4):
-    if schedule.get_jobs():
-        return
-
-    if not hf_users:
-        hf_users = os.getenv("hf_users")
-
-    if not ms_users:
-        ms_users = hf_users
-
-    print(f"监控开启中...每 {period} 小时触发")
-    fixed_activate = partial(activate, hf_users=hf_users, ms_users=ms_users)
-    schedule.every(period).hours.do(fixed_activate)
-    while True:
-        schedule.run_pending()
-        time.sleep(DELAY)
-
-
-def listasks():
-    jobs = schedule.get_jobs()
-    if jobs:
-        details = f"{jobs}".replace("[", "").replace("]", "")
-        return fix_datetime(
-            details.split("functools.")[0] + "(" + details.split(") (")[-1]
-        )
-
-    return "None"
-
-
-with gr.Blocks() as demo:
-    gr.Interface(
-        title="Start keeping all spaces active periodically",
-        fn=monitor,
-        inputs=[
-            gr.Textbox(
-                label="HuggingFace",
-                placeholder="Usernames joint by ;",
-            ),
-            gr.Textbox(
-                label="ModelScope",
-                placeholder="Usernames joint by ;",
-            ),
-        ],
-        outputs=None,
-        allow_flagging="never",
-    )
-    gr.Interface(
-        title="See current task status",
-        fn=listasks,
-        inputs=None,
-        outputs=gr.Textbox(label="Current task details"),
-        allow_flagging="never",
-    )
-    gr.Interface(
-        title="Test activation for all spaces once",
-        fn=activate,
-        inputs=[
-            gr.Textbox(
-                label="HuggingFace",
-                placeholder="Usernames joint by ;",
-            ),
-            gr.Textbox(
-                label="ModelScope",
-                placeholder="Usernames joint by ;",
-            ),
-        ],
-        outputs=gr.Dataframe(label="Activated spaces"),
-        allow_flagging="never",
-    )
 
-demo.launch()
+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��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://www.modelscope.cn/studio/Genius-Society/piano_transcription/resolve/master/MuseScore.AppImage",
+            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 = os.getenv("mscore")
+
+
+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

@@ -1,4 +1,7 @@
-requests
-beautifulsoup4
-tqdm
-schedule
+librosa==0.9.2
+piano_transcription_inference
+pymupdf
+music21
+modelscope[framework]==1.18
+torch
+transformers