AI Paper Atlas

Research map / Detection and segmentation / CNN features

Improving CLIP Fine-tuning Performance

Yixuan Wei, Han Hu, Zhenda Xie, Ze Liu, Zheng Zhang, Yue Cao, Jianmin Bao, Dong Chen, Baining Guo

ICCV 2023 · Detection and segmentation › CNN features · approach: Convolution-transformer backbones · 20 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
feature-distill CLIP ViT-B/16 into a fresh student (FD-CLIP)CLIP ViT-B/16 · ImageNet-1K fine-tuningtop-1 accuracy (%): 82.9 -> 85.0Improved
feature-distill CLIP ViT-B/16CLIP ViT-B/16 · COCOAPbox: 45.0 -> 48.2Improved
feature-distill CLIP ViT-B/16CLIP ViT-B/16 · NYUv2RMSE: 0.416 -> 0.352Improved
use full feature map as distillation target instead of [CLS] token (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 81.9 -> 84.4Improved
use full feature map instead of GAP feature (ablation)FD-CLIP · ADE20KmIoU: 50.3 -> 51.8Improved
distill full images instead of masked partial images (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 84.4 -> 84.4 (75% vs full)No clear effect
standardize teacher feature map instead of none (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 83.5 -> 84.4Improved
standardization instead of l2 norm (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 83.9 -> 84.4Improved
add drop path on teacher branch (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 84.4 -> 84.0Hurt
shared relative position bias instead of APE (ablation)FD-CLIP · ImageNet-1K fine-tuningtop-1 accuracy (%): 84.0 -> 84.4Improved
apply feature distillation to DeiTDeiT ViT-B · NYUv2RMSE: 0.403 -> 0.404No clear effect
apply feature distillation to SwinV2-GSwinV2-G · COCOAPbox: 63.1 -> 64.2Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy