AI Paper Atlas

Research map / Detection and segmentation / CNN features

Embedding-Free Transformer with Inference Spatial Reduction for Efficient Semantic Segmentation

Hyunwoo Yu, Yubin Cho, Beoungwoo Kang, Seunghun Moon, Kyeongbo Kong, Suk-Ju Kang

ECCV 2024 · Detection and segmentation › CNN features · approach: Plain and high-resolution ViTs · 24 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
train with small reduction and apply ISR at inference instead of the same reduction in training (ablation)EDAFormer-T · COCO-StuffmIoU (%): 39.8 -> 40.3Improved
train with small reduction and apply ISR at inference (ablation)EDAFormer-B · COCO-StuffmIoU (%): 45.3 -> 45.8Improved
use embedding-free attention instead of QKV-embedding attention in decoder (ablation)EDAFormer-T · ADE20K valmIoU (%): 41.5 -> 42.3Improved
use global attention instead of depth-wise convolution in decoder (ablation)EDAFormer-T · ADE20K valmIoU (%): 40.7 -> 42.3Improved
allocate more EFT blocks to high-level decoder stages {3-2-1} (ablation)EDAFormer-T · ADE20K valmIoU (%): 41.5 -> 42.3Improved
apply ISR [16,8,2,1]-[2,4,8] at inference without fine-tuningEDAFormer-T · Cityscapes valGFLOPs: 151.7 -> 94.9Improved
apply ISR [16,8,2,1]-[2,4,8] at inferenceEDAFormer-T · ADE20K valmIoU (%): 42.3 -> 42.1No clear effect
push ISR to larger reduction [16,8,4,2]-[2,4,8] without fine-tuning (ablation)EDAFormer-T · COCO-StuffmIoU (%): 40.3 -> 39.1Hurt
fine-tune 40K iterations after larger-reduction ISR (ablation)EDAFormer-T · COCO-StuffmIoU (%): 39.1 -> 40.2Improved
remove QKV embeddings so ISR degrades less at high reduction (ablation)EDAFormer-T decoder with ISR · Cityscapes valmIoU (%): 77.9 -> 78.6Improved
use max pooling instead of average pooling for key-value reduction (ablation)EDAFormer-T with ISR · Cityscapes valmIoU (%): 78.7 -> 78.4Hurt
use bipartite matching instead of average pooling for ISR (ablation)EDAFormer-T with ISR · Cityscapes valFPS (img/s): 10.2 -> 10.5No 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