This is the code implementation for our paper Semi-supervised Concept Bottleneck Models
- Download the dataset (CUB_200_2011.tgz) from https://www.vision.caltech.edu/datasets/cub_200_2011/.
- Unpack CUB_200_2011.tgz to the
data/
directory in this project. - Download the concept annotations (class_attr_data_10) from https://worksheets.codalab.org/bundles/0x5b9d528d2101418b87212db92fea6683, which is from the vanilla CBM.