Noises_Unified_v3 / statistic.py
Prummn's picture
Add files using upload-large-folder tool
efa83a7 verified
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import os
import librosa
from collections import Counter
from tqdm import tqdm
import sys
def is_leaf_dir(path):
"""判断是否为最原子的子文件夹(无子文件夹)"""
for item in os.listdir(path):
if os.path.isdir(os.path.join(path, item)):
return False
return True
def get_wav_length(path):
try:
y, sr = librosa.load(path, sr=None, mono=True)
return round(len(y) / sr)
except Exception as e:
print(f"读取失败: {path} 错误: {e}")
return None
def main(noise_root):
noise_root = os.path.abspath(noise_root)
print("扫描所有 leaf(最原子)子文件夹...\n")
leaf_dirs = []
# 遍历 noise 根目录下所有文件夹
for dataset in os.listdir(noise_root): # MUSAN、DEMAND、ESC50 …
dataset_path = os.path.join(noise_root, dataset)
if not os.path.isdir(dataset_path):
continue
# 深度遍历,找所有最原子的子文件夹
for root, dirs, files in os.walk(dataset_path):
if is_leaf_dir(root):
leaf_dirs.append(root)
print("发现的 leaf 文件夹:")
for d in leaf_dirs:
print(" -", d)
print("\n开始统计...\n")
all_results = {}
for leaf in leaf_dirs:
counter = Counter()
wav_count = 0
# 遍历该 leaf 下的所有 wav
for f in tqdm(os.listdir(leaf), desc=f"[{leaf}]"):
if not f.lower().endswith(".wav"):
continue
wav_count += 1
path = os.path.join(leaf, f)
L = get_wav_length(path)
if L is not None:
counter[L] += 1
all_results[leaf] = (wav_count, counter)
# 打印结果
print("\n============== 统计结果 ==============\n")
for leaf, (cnt, dist) in all_results.items():
print(f"📁 {leaf}")
print(f" WAV 数量:{cnt}")
print(" 时长分布(秒):")
for L in sorted(dist):
print(f" {L:2d}s : {dist[L]}")
print()
print("统计完成!")
if __name__ == "__main__":
if len(sys.argv) < 2:
print("用法: python stats_leaf_folders.py <noise根目录>")
print("示例: python stats_leaf_folders.py noise")
sys.exit(1)
main(sys.argv[1])