AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

FuseTeacher: Modality-fused Encoders are Strong Vision Supervisors

Chen-Wei Xie, Siyang Sun, Liming Zhao, Pandeng Li, Shuailei Ma, Yun Zheng

ECCV 2024 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
add FuseTeacher modality-fused distillation to CLIPCLIP · YFCC15M-Cap avg of 6 ImageNet variants + retrievalaverage (%): 36.4 -> 38.4Improved
add FuseTeacher on top of MaskCLIPMaskCLIP · YFCC15M-Cap avgaverage (%): 38.3 -> 40.3Improved
add classification distillation (ablation)FuseTeacher · ImageNet-1K val zero-shotzero-shot accuracy (%): 41.1 -> 42.9Improved
add retrieval distillation (ablation)FuseTeacher · ImageNet-1K val zero-shotzero-shot accuracy (%): 41.1 -> 43.0Improved
combine classification and retrieval distillation (ablation)FuseTeacher · ImageNet-1K val linear probelinear-probe accuracy (%): 67.6 -> 68.4Improved
use fusion encoder embedding instead of image-only embedding for multimodal queries (ablation)CLIP · ImageNet-MM top-1top-1 accuracy (%): 65.2 -> 78.3Improved
scale backbone from ViT-B/32 to ViT-B/16 with FuseTeacher on Union23M (ablation)FuseTeacher · Union23M avgaverage (%): 53.7 -> 60.6Improved
apply FuseTeacher alone versus MaskCLIP with ViT-L/14FuseTeacher · Union23M ViT-L/14 avgaverage (%): 65.9 -> 64.3Hurt
scale pretraining data from Union23M to Union65M (ablation)FuseTeacher+ ViT-L/14 · average of ImageNet variants + retrievalaverage (%): 67.0 -> 69.9Improved
use FuseTeacher instead of MaskCLIP for COCO image-to-text retrievalFuseTeacher · MS-COCO zero-shot retrievalImage-to-text R@1 (%): 49.4 -> 48.4Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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