ChangeFormer: A Transformer-Based Siamese Network for Change Detection - IGARSS'22
ฝัง
- เผยแพร่เมื่อ 7 ก.พ. 2025
- Video presentation of our IGARSS'22 paper: "ChangeFormer: A Transformer-Based Siamese Network for Change Detection".
Link to paper: arxiv.org/abs/...
Link to code: github.com/wgc...
Project page: www.wgcban.com...
Dear Sir,
i want to train model with multiple class , for example 3.
so mask will be have shape 256*256*3 or 256*256*1.
Hoping for a response.
Thanks
Yes, you can and people have trained the ChangeFormer for multi-class CD. Please see the issues on the github page (github.com/wgcban/ChangeFormer) for more details.
hello sir, I am actually doing a damage assessment project in which I basically segment roads and buildings from satellite images. Can I use this model for pre- and post-disaster image change detection to identify damage?
Thank you
Yes, you can use it. Please refer to the GitHub page to download the pre-trained models.
I converted the code to the Windows version and tested it. If there are two classes, can you add more classes? I want to use it in multiple classification rather than binary classification.
I'm a foreigner, so I apologize for using a translator.
Additionally, I wonder what these two functions do.
multi_scale_train=True
multi_scale_infer=False
@@태민김-z6w Hey yes it supports multiple classes. Please see this issue: github.com/wgcban/ChangeFormer/issues/40
If you have any questions fell free to create an github issue. I will answer those. thanks!