The performance of default voc8.res50v3+.CPS

charlesCXK / TorchSemiSeg

[CVPR 2021] CPS: Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision

MIT License

504 stars 74 forks source link

The performance of default voc8.res50v3+.CPS #39

Open swt199211 opened 3 years ago

swt199211 commented 3 years ago

Hello, I'm training with double 3080 cards. The batchsize can only be set to 4. The learning rate is 0.0025. The ratio of 1 / 8. I've run it three times and the IOU is only 0.66. What's the reason?Does batchsize have a great impact?

charlesCXK commented 3 years ago

Hi, obviously the batch size plays an important role in Semantic Segmentation task. Besides, when the batch size is smaller, the learning rate should be adjusted linearly at the same time

wing212 commented 3 years ago

Hello, I am using two 3090 graphics cards, but I always get an error. The original text uses four graphics cards. What do I need to change in the code to match the two graphics cards?

XieZhiyao21 commented 7 months ago

Have you solved your problem?What changes do you make?