Skip to main content
. Author manuscript; available in PMC: 2024 Dec 5.
Published in final edited form as: Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit. 2024 Sep 16;2024:26140–26150. doi: 10.1109/cvpr52733.2024.02470

Figure 3. Comparison of methods using the CLIP-ResNet50 architecture on four benchmark datasets.

Figure 3.

We use Worst Group Accuracy to evaluate the performance for various methods, including ERM, semi-supervised baselines (i.e., AFR [47], CnC [73], JTT [33]), and our proposed methods. We observe that CFR combined with the sample selection strategies (i.e., {DPS, RPS}×{RNS,NNS}) outperforms all semi-supervised baselines across all benchmarks.