search

ali-vilab / composer

Official implementation of "Composer: Creative and Controllable Image Synthesis with Composable Conditions"
MIT License
1.53k stars 48 forks source link

The dimension of localized conditionings #4

Open jmkim0309 opened 1 year ago

jmkim0309 commented 1 year ago

@huanglianghua In section 2.3,

For localized representations including sketches, segmentation masks, depthmaps, intensity images, and masked images, we project them into uniform dimensional embeddings with the same spatial size as the noisy latent xt using stacked convolutional layers.

What is the value of uniform dimension here?