Image Classification on ImageNet OOD A/R/S variants (test)
59.29Accuracy (ImageNet-A)TPGM-C
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TPGM-CPre-trained=CLIP, Backbone=ViT-L, Protocol=Proposed (Controlled)2023.03 | 59.29 | 76.32 | 65 | 66.87 | 14.91 | |
| TPGMPre-trained=CLIP, Backbone=ViT-L, Protocol=Proposed2023.03 | 58.31 | 74.41 | 65.13 | 65.95 | 13.33 | |
| WISEPre-trained=CLIP, Backbone=ViT-L, Protocol=Weight Interpolation2023.03 | 58.26 | 75.37 | 64.84 | 66.16 | 13.68 | |
| Zero-ShotPre-trained=CLIP, Backbone=ViT-L, Protocol=Zero-Shot2023.03 | 52.21 | 71.69 | 58.24 | 60.71 | 4.33 | |
| Vanilla FTPre-trained=CLIP, Backbone=ViT-L, Protocol=Fine-Tuning2023.03 | 49.67 | 63.29 | 61.62 | 58.19 | 0 |