search

GeWu-Lab / OGM-GE_CVPR2022

The repo for "Balanced Multimodal Learning via On-the-fly Gradient Modulation", CVPR 2022 (ORAL)
MIT License
236 stars 18 forks source link

关于Audio_only和Visual_only准确率的问题 #32

Open yuquanqingyin opened 1 year ago

yuquanqingyin commented 1 year ago

请问论文Table1中的Audio-only和Visual-only是如何得到的？应该不是代码中的acc_a和acc_v吧

echo0409 commented 1 year ago

Hi, thanks for your attention. We indicate the process at the footnote of our paper.

"Here, the visual-only and the audio-only are the models that trained with a single modality. To evaluate the uni-modal encoders of the multimodal model, we observe the performance through fixing the joint-trained uni-modal encoder and finetuning a uni-modal classifier only."

yuquanqingyin commented 1 year ago

能否提供Audio-only和Visual-only的代码和超参数呢？因为我跑出来的结果是：Audio-only：60% Visual-only：20%，与论文Table1中的数据差距较大

HPU-Yz commented 1 year ago

您好，我也是遇到同样的问题，请问下能否提供Audio-only和Visual-only的代码，谢谢

HPU-Yz commented 1 year ago

能否提供Audio-only和Visual-only的代码和超参数呢？因为我跑出来的结果是：Audio-only：60% Visual-only：20%，与论文Table1中的数据差距较大

请问下您有得到吗，或者方便分享下吗，谢谢