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
107 stars 17 forks source link

The testing result is bad #3

Open redizzy opened 4 hours ago

redizzy commented 4 hours ago

hi， I have used best_model and its config on the LRS2 data set provided by you, used the LRS2 data set provided by you and conducted pre-processing, but the test index is poor, as shown in the figure below

d3556df0c8860f6c801d758fc47f3cf
redizzy commented 4 hours ago

could you please provide some advice or help? thanks!

JusperLee commented 2 hours ago

I think it may be that the pre-training of your model is not loading correctly.