跳至内容
菜单
此问题已终结
该问题已被关闭 原因: Question / Code not indented
Manikandan2025年10月25日 05时58分13秒
157 查看

Been working on an image classification model lately - the training accuracy looks great, but the validation accuracy seems to plateau around 88–90%.

I’ve already tried data augmentation and dropout tuning, but the improvement is marginal.

Curious to know - what’s your go-to approach when your model just refuses to generalise better? Do you usually tweak the architecture, try transfer learning, or focus on data cleaning?

形象
丢弃
相关帖文 回复 查看 活动
0
10月 25
2
0
10月 25
10
0
10月 25
9
3
7月 25
1207
0
10月 25
37