MedSegDiff icon indicating copy to clipboard operation
MedSegDiff copied to clipboard

为什么训练出现了完全噪声的结果,如图所示的原图和标签的训练对,20000+图片对(单通道灰度图)训练200epoch得到如图结果,sample_step=30,是否有朋友知道是什么原因呢

Open qxxfd opened this issue 1 year ago • 1 comments

image

qxxfd avatar Nov 30 '24 16:11 qxxfd

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿


Title: Why is the training result completely noisy? As shown in the figure, the original image and label training pair are trained on 20000+ image (single channel) pairs for 200 epoch to obtain the result as shown in the figure, sample_step=30

image

Issues-translate-bot avatar Nov 30 '24 16:11 Issues-translate-bot