app.py

import os
import re
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 *

CACHE_DIR = "./__pycache__"
WEIGHTS_PATH = (
    snapshot_download("monetjoe/CRNN_note_F1_0.9677_pedal_F1_0.9186")
    + "/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.makedirs(cache_dir, exist_ok=True)


def get_audio_file_type(file_path):
    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, save_path):
    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 inference(audio_path: str):
    clean_cache()
    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


def get_first_integer(input_string):
    match = re.search(r"\d+", input_string)
    if match:
        return str(int(match.group()))
    else:
        return ""


def infer(audio_url: str):
    clean_cache()
    download_path = f"{CACHE_DIR}/output"
    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"
            print(audio_url)

        download_path = download_audio(audio_url, download_path)

    midi, title = audio2midi(download_path)
    xml = midi2xml(midi, title)
    mxl = xml2mxl(xml)
    pdf, jpg = xml2jpg(xml)
    abc = xml2abc(xml)

    return download_path, midi, pdf, xml, mxl, abc, jpg


with gr.Blocks() as iface:
    with gr.Tab("Upload mode"):
        gr.Interface(
            fn=inference,
            inputs=gr.Audio(
                label="Please upload 100% of the audio before clicking submit",
                type="filepath",
            ),
            outputs=[
                gr.components.File(label="Download MIDI"),
                gr.components.File(label="Download PDF score"),
                gr.components.File(label="Download MusicXML"),
                gr.components.File(label="Download MXL"),
                gr.Textbox(label="abc notation", show_copy_button=True),
                gr.Image(label="Staff", type="filepath"),
            ],
            allow_flagging="never",
        )

    with gr.Tab("Direct link mode"):
        gr.Interface(
            fn=infer,
            inputs=gr.Textbox(
                label="Input audio direct link URL (Netease Cloud Music can directly input non-VIP song page link automatically resolved)"
            ),
            outputs=[
                gr.Audio(label="Download audio", type="filepath"),
                gr.components.File(label="Download MIDI"),
                gr.components.File(label="Download PDF score"),
                gr.components.File(label="Download MusicXML"),
                gr.components.File(label="Download MXL"),
                gr.Textbox(label="abc notation", show_copy_button=True),
                gr.Image(label="Staff", type="filepath"),
            ],
            allow_flagging="never",
        )

iface.launch()