AI Paper Atlas

Research map / Detection and segmentation / CNN features

Sparse Refinement for Efficient High-Resolution Semantic Segmentation

Zhijian Liu, Zhuoyang Zhang, Samir Khaki, Shang Yang, Haotian Tang, Chenfeng Xu, Kurt Keutzer, Song Han

ECCV 2024 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 4 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
add sparse high-resolution refinement to dense low-res predictionSegFormer-B5 · Cityscapes vallatency (ms): 140.6 -> 38.5 (3.7x), mIoU 81.1 -> 81.2Improved
add SparseRefine to SegNeXt-LSegNeXt-L · Cityscapes valmean IoU: 83.0 -> 82.8Hurt
add SparseRefine to HRNet-W48 vs dense half-resolutionHRNet-W48 · Cityscapes valmean IoU: 79.2 -> 80.9Improved
use SparseRefine instead of PointRend refinementHRNet-W48 · Cityscapes valmean IoU: 79.9 -> 80.9Improved
use sparse pixel refinement instead of patch refinementHRNet-W48 · Cityscapes vallatency (ms): 44.1 -> 32.4Improved
use entropy selector instead of random selector (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 79.2 -> 80.9Improved
use entropy selector instead of magnitude selector (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 80.2 -> 80.9Improved
use entropy selector instead of learnable selector (ablation)SparseRefine HRNet-W48 · Cityscapes valselector latency: 4.0ms -> 2.3msImproved
add gated ensembler instead of entropy ensembler (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 80.3 -> 80.9Improved
add logits or features as input to sparse extractor (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 80.9 -> 80.2Hurt
refine all pixels densely instead of sparse 11.8% (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 80.9 -> 80.5Hurt
lower entropy threshold alpha from 0.8 to 0.1 (more pixels refined) (ablation)SparseRefine HRNet-W48 · Cityscapes valmIoU: 79.9 -> 81.1 (latency 24.6 ms -> 40.1 ms)Improved

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