search

wgcban / ddpm-cd

Remote Sensing Change Detection using Denoising Diffusion Probabilistic Models
https://www.wgcban.com/research#h.ar24vwqlm021
MIT License
247 stars 29 forks source link

about fe_A_t, fd_A_t #18

Closed Cassiatora closed 1 year ago

Cassiatora commented 1 year ago

Hi, Thank you for your great work. I want to konw the meaning of fe_A_t, fd_A_t, and opt['model_cd']['feat_type'].

wgcban commented 1 year ago

Hi @Cassiatora

fe_A_t and fd_A_t denote the encoder and decoder feature representations of pre-change image (A) for timestep t. Similarly, fe_B_t, and fe_B_t denote the encoder and decoder feature representations of post-change image (B) for timestep t.

defined as in here: https://github.com/wgcban/ddpm-cd/blob/b0213c0049bab215e470326d97499ae69416a843/model/model.py#L84-L92

wgcban commented 1 year ago

The opt['model_cd']['feat_type'] specifies whether you want encoder (enc) or decoder (dec) features to consider for the change detection. We observed better performance with decoder features. So, default configuration is dec as defined here: https://github.com/wgcban/ddpm-cd/blob/b0213c0049bab215e470326d97499ae69416a843/config/dsifn.json#L62

wgcban commented 1 year ago

@Cassiatora Please let me know if you need more clarification on it. Thanks.

Cassiatora commented 1 year ago

OK, I got it. Your clarification has been very detailed. Thank you for your patience.