search

JusperLee / IIANet

This is the demo of our paper "IIANet: An Intra- and Inter-Modality Attention Network for Audio-Visual Speech Separation".
Apache License 2.0
111 stars 17 forks source link

inference.py Audio results are weird. #4

Open d0v0h opened 3 days ago

d0v0h commented 3 days ago

When I run inference.py , the results for video_tracked1.mp4 and video_trcked2.mp4 are good, but the audio output results are weird. The two audio output results are almost identical.

Is it correct to download the model from the path below?

lrw_resnet18_mstcn.pth.tar : https://drive.google.com/file/d/1RSV4nr0C7MxDWH0FnNyR8ldjjtPOiZFq/view checkpoints/vox2/ : https://github.com/JusperLee/IIANet/releases/download/v1.0.0/vox2.zip

Thank you for reading my post.

world2025 commented 7 hours ago

@JusperLee 请问一下，该inference.py支持分别上传音频和视频吗

world2025 commented 7 hours ago

@JusperLee 请问一下，该inference.py支持分别上传音频和视频吗

而且是实时的推理音频和视频

world2025 commented 7 hours ago

@JusperLee 请问一下，该inference.py支持分别上传音频和视频吗

而且是实时的推理音频和视频

还就是想问下目前只支持2个人的视频吗

d0v0h commented 7 hours ago

你好，虽然这是一个与问题无关的内容，但请问 inference.py 文件的执行结果正常吗？