AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Grid-Attention: Enhancing Computational Efficiency of Large Vision Models without Fine-Tuning

Pengyu Li, biao wang, Tianchu Guo, Xian-Sheng Hua

ECCV 2024 · Architectures and efficiency › Pruning and efficient tuning · approach: Efficient ViT adaptation · 1 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
plug GridAttn into frozen SAM-B without fine-tuningSAM-B · COCO instance segmentationAP: 41.02 -> 41.39 AP; GFLOPs 969.6 -> 813.8Improved
plug GridAttn into frozen SAM-H without fine-tuningSAM-H · COCO instance segmentationAP: 46.13 -> 46.12 AP; GFLOPs 5952 -> 5692No clear effect
use Expedit-SAM token merging instead of GridAttnSAM-H · COCO instance segmentationAP: 46.13 -> 42.33Hurt
add GridAttn on top of Expedit-SAMExpedit-SAM · COCO instance segmentation (SAM-H)GFLOPs / AP: 5094 -> 4893 GFLOPs; AP 42.33 -> 42.42Improved
naive 2x2 sub-image sampling (lower boundary) instead of GridAttn (ablation)frozen SAM-B · COCO instance segmentationAP: 41.39 -> 36.50Hurt
fine-tune SegFormer with GridAttn for 10% iterations instead of naive fine-tuningSegFormer B0-B5 · ADE20KmIoU (MS): B1: 43.9 -> 44.1; B4: 51.0 -> 51.3Improved
use Shuffle strategy instead of Straightforward when training from scratch (ablation)GridAttn SegFormer · ADE20K from scratchmIoU (MS): B4: 49.1 -> 50.7; B5: 51.1 -> 51.8Improved
use Deformable strategy instead of Straightforward when training from scratch (ablation)GridAttn SegFormer-B5 · ADE20K from scratchmIoU (MS): 51.1 -> 52.0; 423.7 -> 305.4 GFLOPs vs benchmarkImproved
use Deformable strategy instead of Shuffle (ablation)GridAttn SegFormer-B5 · ADE20K from scratchGPU latency: 56.2 -> 240 msHurt
increase grid stride S (ablation)frozen SAM with GridAttn · COCO instance segmentationperformance-efficiency tradeoffMixed
reduce fine-tuning iterations to 1600 (1%) (ablation)GridAttn SegFormer · ADE20KmIoU (MS): B3: 50.4 -> 49.5Hurt
fine-tune more than 8000 iterations (ablation)GridAttn SegFormer · ADE20KmIoUNo clear effect

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