zyh-uaiaaaa

Hi, we carry out experiments on FERPlus with 7 basic expression classes, same as RAF-DB. As for AffectNet, we conduct pre-processing on the data and use a balanced sampler during...

Hi. We think that completely reproducing the results are not guaranteed across PyTorch releases, different platforms or CPU/GPU executions, even when using identical seeds.

Hi Amazedan, Firstly, you can assess the performance of the trained model by evaluating it on all test samples and obtaining the loss values for each sample. Afterward, you can...

学习率需要从1e-4调整到2e-4。 这个不太可能，估计是你代码某个地方出了问题，或者标签没有读对，82%的准确率过于低了。 你可以参考一下我对其他人问题的回复，如下： “I have re-run my code without making any changes. Using the pre-trained ResNet-50 model, it easily achieved an accuracy of 90.35% and the best accuracy recorded was...

It seems the attentive center loss branch has little impact on the accuracy.