AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Stitched ViTs are Flexible Vision Backbones

Zizheng Pan, Jing Liu, Haoyu He, Jianfei Cai, Bohan Zhuang

ECCV 2024 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 3 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 Two-way Stitching (TWS) to SN-Netv1 (ablation)SN-Netv1 (DeiT3-S/L) · ADE20K, 34G FLOPsmIoU (%): 40.4 -> 42.8Improved
add Two-way Stitching (TWS) to SN-Netv1 (ablation)SN-Netv1 (DeiT3-S/L) · ADE20K, 363G FLOPsmIoU (%): 51.6 -> 50.5Hurt
add Resource-constrained Sampling (ROS) on top of TWS (ablation)SN-Netv2 (DeiT3-S/L) · ADE20K, 34G FLOPsmIoU (%): 42.8 -> 43.7Improved
add LoRA on stitching layers on top of TWS and ROS (ablation)SN-Netv2 (DeiT3-S/L) · ADE20K, 34G FLOPsmIoU (%): 43.7 -> 45.6Improved
enlarge stitching space with TWS alone on ImageNet-1K (ablation)SN-Netv1 (DeiT3-S/L) · ImageNet-1KTop-1 accuracy (%): -17.2% Top-1 at small anchorHurt
use SN-Netv2 instead of training individual anchors separatelyDeiT3-B/L on ADE20K · ADE20Ktraining cost (A100 GPU hours): 140 vs 174 + 90Improved
use SN-Netv2 with LoRA on backbone instead of DeiT3-L with LoRACOCO-Stuff-10K backbone PEFT · COCO-Stuff-10KmIoU (%): 44.8 -> 45.8Improved
apply LoRA PEFT to the entire backbone instead of full finetuning (ablation)SN-Netv2 · COCO-Stuff-10KmIoU (%): 48.2 -> 45.8 (top end)Hurt
train SN-Netv2 on detection with heavy ViTDet decoderSN-Netv2 DeiT3-S/L · COCO-2017 detectionbbox AP: 51.2 vs 51.9 bbox AP at DeiT3-L FLOPsMixed
remove low-rank update and fully finetune stitching layers (ablation)SN-Netv2 (DeiT3-S/L) · COCO-Stuff-10KmIoUHurt

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