Training CNN on training set with different numbers of images in the training set

2 次查看（过去 30 天）

显示更早的评论

Andrea Giometto 2021-3-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/776697-training-cnn-on-training-set-with-different-numbers-of-images-in-the-training-set

回答： Srivardhan Gadila 2021-3-23

采纳的回答： Srivardhan Gadila

Hi, I'm doing transfer learning to train squeezenet on a database of microcopy images. I have two classes that have different numbers of images (class 1: 200, class2: 400). Is there a way in Matlab to assign lower weight to the images in class 2, so that the classification is not biased towards the most numerous class? Thank you

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Srivardhan Gadila 2021-3-23

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/776697-training-cnn-on-training-set-with-different-numbers-of-images-in-the-training-set#answer_655362

You can enable ClassWeights for classificationLayer or you can also use focalLossLayer.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

更多回答（0 个）

请先登录，再回答此问题。

类别

AI and Statistics Deep Learning Toolbox

在 Help Center 和 File Exchange 中查找有关 Deep Learning Toolbox 的更多信息

标签

产品

MATLAB

版本

R2020b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Translated by