Open dwangxxx opened 4 years ago
Hello, sorry to bother again. In this paper, you show the results finetuned on the COCOstuff dataset. I want to know how to finetune the model on the COCOstuff dataset for ImageNet dataset. I don't know if I'm right, but I thought of the following steps: (1) Train the model on the ImageNet dataset. (2) Finetune the model from (1) on the COCOStuff dataset with a smaller learning rate. Maybe I was wrong, but can you provide me some detailed information, such as finetune learning rate and some other parameters?
Hello, I've read your paper recently. In this paper, you mentioned that the training epoch is set as 2, 5 and 2, but the source code is set 150. I want to know which is correct? Thank you!