search

IDEA-Research / MP-Former

[CVPR 2023] Official implementation of the paper: MP-Former: Mask-Piloted Transformer for Image Segmentation
Other
118 stars 6 forks source link

About mask2former_transformer_decoder.py #7

Closed HarisIqbal89 closed 10 months ago

HarisIqbal89 commented 10 months ago

hello author, may I ask a question about mask2former_transformer_decoder.py. In line 719~728，the code get some parameters from config. Such as dn_mode and head_dn，But I could not find them among the config. So, I want to know whether it matter to the result？Actually, I want to change my code which based mask2former to MP-former. I simply change the mask2former_transformer_decoder.py and criterion.py. Do I need to change any other code, I am looking forward to receive reply from you :) ret["mask_dim"] = cfg.MODEL.SEM_SEG_HEAD.MASK_DIM ret["dn_mode"] = cfg.MODEL.MASK_FORMER.DN_MODE ret["head_dn"] = cfg.MODEL.MASK_FORMER.HEAD_DN ret["all_lys"] = cfg.MODEL.MASK_FORMER.ALL_LY_DN ret["dn_ratio"] = cfg.MODEL.MASK_FORMER.DN_RATIO ret["dn_label_noise_ratio"] = cfg.MODEL.MASK_FORMER.LB_NOISE_RATIO

HaoZhang534 commented 10 months ago

Some parameters are defined in link , you can overwrite them in config file or in command line. If you do not overwrite them, they will use the default value in https://github.com/IDEA-Research/MP-Former/blob/a71f504a467d7079cf02726a05ff23bbb933607d/mask2former/config.py

HarisIqbal89 commented 10 months ago

Some parameters are defined in link , you can overwrite them in config file or in command line. If you do not overwrite them, they will use the default value in https://github.com/IDEA-Research/MP-Former/blob/a71f504a467d7079cf02726a05ff23bbb933607d/mask2former/config.py

thank you for your reply.So I just need to change three codes to modify mask2former to MPformer: mask2former_transformer_decoder.py and criterion.py and config.py，right？MP-former is wonderful！