Spaces:
Running
Running
admin
commited on
Commit
·
a0eed22
1
Parent(s):
f398992
sync ms
Browse files
app.py
CHANGED
|
@@ -9,7 +9,7 @@ import librosa.display
|
|
| 9 |
import matplotlib.pyplot as plt
|
| 10 |
from collections import Counter
|
| 11 |
from model import EvalNet
|
| 12 |
-
from utils import get_modelist, find_files, embed_img
|
| 13 |
|
| 14 |
|
| 15 |
TRANSLATE = {
|
|
@@ -344,33 +344,38 @@ def most_frequent_value(lst: list):
|
|
| 344 |
|
| 345 |
|
| 346 |
def infer(wav_path: str, log_name: str, folder_path=TEMP_DIR):
|
| 347 |
-
|
| 348 |
-
|
|
|
|
|
|
|
|
|
|
| 349 |
|
| 350 |
-
|
| 351 |
-
|
| 352 |
|
| 353 |
-
|
| 354 |
-
|
| 355 |
-
try:
|
| 356 |
model = EvalNet(log_name, len(TRANSLATE)).model
|
| 357 |
eval("wav2%s" % spec)(wav_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 358 |
|
| 359 |
-
|
| 360 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 361 |
|
| 362 |
-
|
| 363 |
-
|
| 364 |
-
for jpg in jpgs:
|
| 365 |
-
input = embed_img(jpg)
|
| 366 |
-
output: torch.Tensor = model(input)
|
| 367 |
-
preds.append(torch.max(output.data, 1)[1])
|
| 368 |
|
| 369 |
-
|
| 370 |
-
return (
|
| 371 |
-
os.path.basename(wav_path),
|
| 372 |
-
f"{TRANSLATE[CLASSES[pred_id]][0]} ({TRANSLATE[CLASSES[pred_id]][1].capitalize()})",
|
| 373 |
-
)
|
| 374 |
|
| 375 |
|
| 376 |
if __name__ == "__main__":
|
|
@@ -385,39 +390,40 @@ if __name__ == "__main__":
|
|
| 385 |
gr.Interface(
|
| 386 |
fn=infer,
|
| 387 |
inputs=[
|
| 388 |
-
gr.Audio(label="
|
| 389 |
-
gr.Dropdown(choices=models, label="
|
| 390 |
],
|
| 391 |
outputs=[
|
| 392 |
-
gr.Textbox(label="
|
|
|
|
| 393 |
gr.Textbox(
|
| 394 |
-
label="
|
| 395 |
show_copy_button=True,
|
| 396 |
),
|
| 397 |
],
|
| 398 |
examples=examples,
|
| 399 |
cache_examples=False,
|
| 400 |
flagging_mode="never",
|
| 401 |
-
title="
|
| 402 |
)
|
| 403 |
|
| 404 |
gr.Markdown(
|
| 405 |
-
""
|
| 406 |
-
|
| 407 |
-
```bibtex
|
| 408 |
-
@article{Zhou-2025,
|
| 409 |
-
|
| 410 |
-
|
| 411 |
-
|
| 412 |
-
|
| 413 |
-
|
| 414 |
-
|
| 415 |
-
|
| 416 |
-
|
| 417 |
-
|
| 418 |
-
|
| 419 |
-
}
|
| 420 |
-
```"""
|
| 421 |
)
|
| 422 |
|
| 423 |
demo.launch()
|
|
|
|
| 9 |
import matplotlib.pyplot as plt
|
| 10 |
from collections import Counter
|
| 11 |
from model import EvalNet
|
| 12 |
+
from utils import get_modelist, find_files, embed_img, _L, EN_US
|
| 13 |
|
| 14 |
|
| 15 |
TRANSLATE = {
|
|
|
|
| 344 |
|
| 345 |
|
| 346 |
def infer(wav_path: str, log_name: str, folder_path=TEMP_DIR):
|
| 347 |
+
status = "Success"
|
| 348 |
+
filename = result = None
|
| 349 |
+
try:
|
| 350 |
+
if os.path.exists(folder_path):
|
| 351 |
+
shutil.rmtree(folder_path)
|
| 352 |
|
| 353 |
+
if not wav_path:
|
| 354 |
+
return None, "请输入音频!"
|
| 355 |
|
| 356 |
+
spec = log_name.split("_")[-3]
|
| 357 |
+
os.makedirs(folder_path, exist_ok=True)
|
|
|
|
| 358 |
model = EvalNet(log_name, len(TRANSLATE)).model
|
| 359 |
eval("wav2%s" % spec)(wav_path)
|
| 360 |
+
jpgs = find_files(folder_path, ".jpg")
|
| 361 |
+
preds = []
|
| 362 |
+
for jpg in jpgs:
|
| 363 |
+
input = embed_img(jpg)
|
| 364 |
+
output: torch.Tensor = model(input)
|
| 365 |
+
preds.append(torch.max(output.data, 1)[1])
|
| 366 |
|
| 367 |
+
pred_id = most_frequent_value(preds)
|
| 368 |
+
filename = os.path.basename(wav_path)
|
| 369 |
+
result = (
|
| 370 |
+
TRANSLATE[CLASSES[pred_id]][1].capitalize()
|
| 371 |
+
if EN_US
|
| 372 |
+
else f"{TRANSLATE[CLASSES[pred_id]][0]} ({TRANSLATE[CLASSES[pred_id]][1].capitalize()})"
|
| 373 |
+
)
|
| 374 |
|
| 375 |
+
except Exception as e:
|
| 376 |
+
status = f"{e}"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 377 |
|
| 378 |
+
return status, filename, result
|
|
|
|
|
|
|
|
|
|
|
|
|
| 379 |
|
| 380 |
|
| 381 |
if __name__ == "__main__":
|
|
|
|
| 390 |
gr.Interface(
|
| 391 |
fn=infer,
|
| 392 |
inputs=[
|
| 393 |
+
gr.Audio(label=_L("上传录音"), type="filepath"),
|
| 394 |
+
gr.Dropdown(choices=models, label=_L("选择模型"), value=models[0]),
|
| 395 |
],
|
| 396 |
outputs=[
|
| 397 |
+
gr.Textbox(label=_L("状态栏"), show_copy_button=True),
|
| 398 |
+
gr.Textbox(label=_L("音频文件名"), show_copy_button=True),
|
| 399 |
gr.Textbox(
|
| 400 |
+
label=_L("中国乐器识别"),
|
| 401 |
show_copy_button=True,
|
| 402 |
),
|
| 403 |
],
|
| 404 |
examples=examples,
|
| 405 |
cache_examples=False,
|
| 406 |
flagging_mode="never",
|
| 407 |
+
title=_L("建议录音时长保持在 3s 左右"),
|
| 408 |
)
|
| 409 |
|
| 410 |
gr.Markdown(
|
| 411 |
+
f"# {_L('引用')}"
|
| 412 |
+
+ """
|
| 413 |
+
```bibtex
|
| 414 |
+
@article{Zhou-2025,
|
| 415 |
+
author = {Monan Zhou and Shenyang Xu and Zhaorui Liu and Zhaowen Wang and Feng Yu and Wei Li and Baoqiang Han},
|
| 416 |
+
title = {CCMusic: An Open and Diverse Database for Chinese Music Information Retrieval Research},
|
| 417 |
+
journal = {Transactions of the International Society for Music Information Retrieval},
|
| 418 |
+
volume = {8},
|
| 419 |
+
number = {1},
|
| 420 |
+
pages = {22--38},
|
| 421 |
+
month = {Mar},
|
| 422 |
+
year = {2025},
|
| 423 |
+
url = {https://doi.org/10.5334/tismir.194},
|
| 424 |
+
doi = {10.5334/tismir.194}
|
| 425 |
+
}
|
| 426 |
+
```"""
|
| 427 |
)
|
| 428 |
|
| 429 |
demo.launch()
|
model.py
CHANGED
|
@@ -1,8 +1,9 @@
|
|
| 1 |
import torch
|
| 2 |
import torch.nn as nn
|
| 3 |
import torchvision.models as models
|
|
|
|
| 4 |
from datasets import load_dataset
|
| 5 |
-
from utils import MODEL_DIR
|
| 6 |
|
| 7 |
|
| 8 |
class EvalNet:
|
|
@@ -17,7 +18,7 @@ class EvalNet:
|
|
| 17 |
self.m_type, self.input_size = self._model_info(m_ver)
|
| 18 |
|
| 19 |
if not hasattr(models, m_ver):
|
| 20 |
-
raise
|
| 21 |
|
| 22 |
self.model = eval("models.%s()" % m_ver)
|
| 23 |
linear_output = self._set_outsize()
|
|
@@ -34,11 +35,15 @@ class EvalNet:
|
|
| 34 |
if ver == bb["ver"]:
|
| 35 |
return bb
|
| 36 |
|
| 37 |
-
print("
|
| 38 |
return backbone_list[0]
|
| 39 |
|
| 40 |
def _model_info(self, m_ver: str):
|
| 41 |
-
backbone_list =
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
backbone = self._get_backbone(m_ver, backbone_list)
|
| 43 |
m_type = str(backbone["type"])
|
| 44 |
input_size = int(backbone["input_size"])
|
|
|
|
| 1 |
import torch
|
| 2 |
import torch.nn as nn
|
| 3 |
import torchvision.models as models
|
| 4 |
+
from modelscope.msdatasets import MsDataset
|
| 5 |
from datasets import load_dataset
|
| 6 |
+
from utils import MODEL_DIR, EN_US
|
| 7 |
|
| 8 |
|
| 9 |
class EvalNet:
|
|
|
|
| 18 |
self.m_type, self.input_size = self._model_info(m_ver)
|
| 19 |
|
| 20 |
if not hasattr(models, m_ver):
|
| 21 |
+
raise ValueError("不支持的模型")
|
| 22 |
|
| 23 |
self.model = eval("models.%s()" % m_ver)
|
| 24 |
linear_output = self._set_outsize()
|
|
|
|
| 35 |
if ver == bb["ver"]:
|
| 36 |
return bb
|
| 37 |
|
| 38 |
+
print("未找到骨干网络名称,使用默认选项 - alexnet")
|
| 39 |
return backbone_list[0]
|
| 40 |
|
| 41 |
def _model_info(self, m_ver: str):
|
| 42 |
+
backbone_list = (
|
| 43 |
+
load_dataset("monetjoe/cv_backbones", split="train")
|
| 44 |
+
if EN_US
|
| 45 |
+
else MsDataset.load("monetjoe/cv_backbones", split="v1")
|
| 46 |
+
)
|
| 47 |
backbone = self._get_backbone(m_ver, backbone_list)
|
| 48 |
m_type = str(backbone["type"])
|
| 49 |
input_size = int(backbone["input_size"])
|
requirements.txt
CHANGED
|
@@ -1,5 +1,7 @@
|
|
| 1 |
-
torch
|
| 2 |
-
|
|
|
|
|
|
|
| 3 |
librosa
|
| 4 |
matplotlib
|
| 5 |
-
|
|
|
|
| 1 |
+
torch==2.6.0+cu118
|
| 2 |
+
-f https://download.pytorch.org/whl/torch
|
| 3 |
+
torchvision==0.21.0+cu118
|
| 4 |
+
-f https://download.pytorch.org/whl/torchvision
|
| 5 |
librosa
|
| 6 |
matplotlib
|
| 7 |
+
modelscope[framework]==1.21.0
|
utils.py
CHANGED
|
@@ -1,10 +1,37 @@
|
|
| 1 |
import os
|
| 2 |
import torch
|
| 3 |
import torchvision.transforms as transforms
|
| 4 |
-
|
|
|
|
| 5 |
from PIL import Image
|
| 6 |
|
| 7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
|
| 10 |
def toCUDA(x):
|
|
@@ -27,19 +54,16 @@ def find_files(folder_path=f"{MODEL_DIR}/examples", ext=".wav"):
|
|
| 27 |
|
| 28 |
|
| 29 |
def get_modelist(model_dir=MODEL_DIR, assign_model=""):
|
| 30 |
-
try:
|
| 31 |
-
entries = os.listdir(model_dir)
|
| 32 |
-
except OSError as e:
|
| 33 |
-
print(f"Cannot access {model_dir}: {e}")
|
| 34 |
-
return
|
| 35 |
-
|
| 36 |
output = []
|
| 37 |
-
for entry in
|
|
|
|
| 38 |
full_path = os.path.join(model_dir, entry)
|
|
|
|
| 39 |
if entry == ".git" or entry == "examples":
|
| 40 |
-
print(f"
|
| 41 |
continue
|
| 42 |
|
|
|
|
| 43 |
if os.path.isdir(full_path):
|
| 44 |
model = os.path.basename(full_path)
|
| 45 |
if assign_model and assign_model.lower() in model:
|
|
|
|
| 1 |
import os
|
| 2 |
import torch
|
| 3 |
import torchvision.transforms as transforms
|
| 4 |
+
import huggingface_hub
|
| 5 |
+
import modelscope
|
| 6 |
from PIL import Image
|
| 7 |
|
| 8 |
+
EN_US = os.getenv("LANG") != "zh_CN.UTF-8"
|
| 9 |
+
|
| 10 |
+
ZH2EN = {
|
| 11 |
+
"上传录音": "Upload a recording",
|
| 12 |
+
"选择模型": "Select a model",
|
| 13 |
+
"状态栏": "Status",
|
| 14 |
+
"音频文件名": "Audio filename",
|
| 15 |
+
"中国乐器识别": "Chinese instrument recognition",
|
| 16 |
+
"建议录音时长保持在 3s 左右": "It is recommended to keep the recording length around 3s.",
|
| 17 |
+
"引用": "Cite",
|
| 18 |
+
}
|
| 19 |
+
|
| 20 |
+
MODEL_DIR = (
|
| 21 |
+
huggingface_hub.snapshot_download(
|
| 22 |
+
"ccmusic-database/CTIS",
|
| 23 |
+
cache_dir="./__pycache__",
|
| 24 |
+
)
|
| 25 |
+
if EN_US
|
| 26 |
+
else modelscope.snapshot_download(
|
| 27 |
+
"ccmusic-database/CTIS",
|
| 28 |
+
cache_dir="./__pycache__",
|
| 29 |
+
)
|
| 30 |
+
)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
def _L(zh_txt: str):
|
| 34 |
+
return ZH2EN[zh_txt] if EN_US else zh_txt
|
| 35 |
|
| 36 |
|
| 37 |
def toCUDA(x):
|
|
|
|
| 54 |
|
| 55 |
|
| 56 |
def get_modelist(model_dir=MODEL_DIR, assign_model=""):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
output = []
|
| 58 |
+
for entry in os.listdir(model_dir):
|
| 59 |
+
# 获取完整路径
|
| 60 |
full_path = os.path.join(model_dir, entry)
|
| 61 |
+
# 跳过'.git'文件夹
|
| 62 |
if entry == ".git" or entry == "examples":
|
| 63 |
+
print(f"跳过 .git 或 examples 文件夹: {full_path}")
|
| 64 |
continue
|
| 65 |
|
| 66 |
+
# 检查条目是文件还是目录
|
| 67 |
if os.path.isdir(full_path):
|
| 68 |
model = os.path.basename(full_path)
|
| 69 |
if assign_model and assign_model.lower() in model:
|