search

NaJaeMin92 / FixBi

Official code for the CVPR 2021 paper "FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation"
63 stars 11 forks source link

how to train when the source dataset has labels but the target doesn't #8

Closed rotem154154 closed 2 years ago

rotem154154 commented 2 years ago

i'm trying to transfer dataset that has labels to simular dataset without labels like in this example https://www.youtube.com/watch?v=9d4jmPmTWmc&t=570s Screen Shot 2021-10-15 at 13 59 49 from the code that load the datasets its look like both the source and target dataset supposed to have labels. https://github.com/NaJaeMin92/FixBi/blob/3c986946fd1582cc0e59510115d5c2e1d307b412/src/dataset.py#L23

NaJaeMin92 commented 2 years ago

Hi @rotem154154

There's no reason to use an ImageFolder if you don't have any labels at all. ImageFolder returns samples and labels according to the folder of each class. I used ImageFolder to access the source labels.