search

nchucvml / STVT

Video Summarization With Spatiotemporal Vision Transformer
Apache License 2.0
17 stars 7 forks source link

Hyperparameter tuning #2

Closed kiss-maker closed 3 months ago

kiss-maker commented 4 months ago

請問您在SumMe數據集上訓練時的超參數是怎樣設置的，我按照您的方法訓練指標只有38。Could you please let me know how you set the hyperparameters during training on the SumMe dataset? I followed your method, but the performance metrics only reached 38.

nchucvml commented 4 months ago

We are not sure about your settings. If you can test the result on the TVSum dataset successfully, the program should be fine. Different random splits might also affect the results.