Thanks for the earlier answer. Do you have any previous model test results of training directly in ADE20K without using classification pre-training?

cheerss / CrossFormer

The official code for the paper: https://openreview.net/forum?id=_PHymLIxuI

MIT License

360 stars 43 forks source link

Thanks for the earlier answer. Do you have any previous model test results of training directly in ADE20K without using classification pre-training? #15

Open 754467737 opened 1 year ago

754467737 commented 1 year ago

I'm using your model directly, with the configuration unchanged, just setting the pre-trained model to None. Do you think the 80000iter training is reasonable?