search

YuanGongND / ast

Code for the Interspeech 2021 paper "AST: Audio Spectrogram Transformer".
BSD 3-Clause "New" or "Revised" License
1.06k stars 203 forks source link

In my own dataset, why is the Avg precision always 0.5 in each epoch? #95

Open 1244547821 opened 1 year ago

1244547821 commented 1 year ago

Hi yuan, in my own two-category data set training, the Avg precision in each epoch is 0.5, Recall is always 1, can you answer my question, the following is my result. start validation acc: 0.934211 AUC: 0.981920 Avg Precision: 0.500000 Avg Recall: 1.000000 d_prime: 2.962967 train_loss: 0.260971 valid_loss: 0.466797 validation finished Epoch-2 lr: 1e-05

YuanGongND commented 1 year ago

Have you tried other baselines? It seems the model is biased to one class (i.e., predict all samples as that class), which could be due to the model is not well-trained, or the data has some problem.