search

MengboLi / MS-SENet

7 stars 1 forks source link

The reproduction results are quite poor. #2

Open wyy0926 opened 1 month ago

wyy0926 commented 1 month ago

I achieved a weighted recall of about 87% when replicating on the RAVDESS dataset, which falls short of the 93.28% reported in the paper. Here are my training parameters： python MS-SENet.py --mode train --data RAVDE --split_fold 10 --random_seed 46 --epoch 200 --gpu 0

MengboLi commented 1 month ago

Thank you for your feedback! I have recently uploaded the model I used. The network structure and parameters may need adjustments depending on the dataset to achieve optimal results. I will gradually organize and share these details to assist others in better replicating the results. If you have any further questions or suggestions, please feel free to discuss them!