Downloads · 30 days
0
DaVinci-2nd/LostEcho
LostEcho is a machine learning model from DaVinci-2nd. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
基于频谱域 U-Net + PatchGAN 的音频修复模型(回声/噪声消除)。模型在幅度谱上预测残差,配合重叠分块 + Hann 窗加权合成,支持任意长度音频。
Downloads · 30 days
0
Access
Public
Updated Aug 30, 2026
Repo size
1.1 GB
Likes
0
Public
Click a slice to open those files.
.pth1.1 GB · 100%
From the Hugging Face model README
基于频谱域 U-Net + PatchGAN 的音频修复模型(回声/噪声消除)。模型在幅度谱上预测残差,配合重叠分块 + Hann 窗加权合成,支持任意长度音频。
| 文件 | 说明 |
|---|---|
lostecho-v1.pth | 最新版本(推荐) |
lostecho-v0.pth | 早期版本 |
import torch
from model import SpectralUNet
model = SpectralUNet(init_channels=48, n_down=4, in_channels=2, out_channels=2, reduction=16)
ckpt = torch.load("lostecho-v1.pth", map_location="cpu")
state = ckpt["model_state_dict"] if "model_state_dict" in ckpt else ckpt
model.load_state_dict(state)
命令行推理(配合本仓库代码):
python inference.py -i input.wav -o output.wav -c lostecho-v1.pth