search

shahaf1313 / ProCST

Official implementation of ProCST image-to-image translation for UDA-SS.
MIT License
28 stars 3 forks source link

Why is the model deeplabv2? #7

Open juajua56 opened 1 year ago

juajua56 commented 1 year ago

Can't we train with HRDA + ProCST model?

shahaf1313 commented 12 months ago

Hey juajua :) We indeed haven't assembled to our architecture more accurate segmentation models, such as DAFormer and HRDA. It can potentially upgrade performance, but it also requires an enormous amount of resource usage such as RAM memory, and thus restricts the maximum output image size. Low resolution images are insufficient to achive improved performance on the UDA-SS task, therefore we haven't conducted those experiments. Best regards, Shahaf