AI Paper Atlas

Research map / Detection and segmentation / CNN features

HyPiDecoder: Hybrid Pixel Decoder for Efficient Segmentation and Detection

Fengzhe Zhou, Humphrey Shi

ICCV 2025 · Detection and segmentation › CNN features · approach: Boundary-aware instance segmentation · 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
replace part of MSDeformAttn pixel decoder with HyPiDecoder (3d3r)Mask DINO SwinL · COCO panoptic val2017PQ: 58.3 -> 58.8Improved
replace default pixel decoder with HyPiDecoder (3d3s)Mask2Former SwinL · COCO panoptic val2017PQ: 57.6 -> 58.1Improved
replace default pixel decoder with HyPiDecoder (3d3s)Mask2Former R50 · COCO panoptic val2017PQ: 47.0 -> 47.5Improved
replace MSDeformAttn with MSLinearAttn plus FPN layers (3l3r)Mask DINO R50 · COCO panoptic val2017PQ: 53.0 -> 53.4Improved
HyPiDecoder 3d3r instead of default pixel decoderMask DINO SwinL · COCO instance segmentation val2017Mask AP: 52.3 -> 52.3No clear effect
HyPiDecoder 3d3s instead of default pixel decoderMask2Former SwinL · COCO instance segmentation val2017Mask AP: 49.5 -> 49.9Improved
HyPiDecoder 3d3s instead of default pixel decoderMask2Former R50 · ADE20K valmIoU: 47.2 -> 47.6Improved
HyPiDecoder 3d3s instead of default pixel decoderMask2Former R50 · Cityscapes valmIoU: 79.4 -> 81.1Improved
add more layers (1l1r to 3l3r) (ablation)HyPiDecoder Mask DINO R50 · COCO panoptic val2017PQ: 50.5 -> 53.4Improved
use MSLinearAttn instead of MSDeformAttn in hybrid layers (ablation)HyPiDecoder Mask DINO R50 (12 ep) · COCO panoptic val2017PQ: 49.0 -> 49.3No clear effect
mix attention and FPN layers (3l3r) instead of linear attention only (6l0r) (ablation)HyPiDecoder Mask DINO R50 (12 ep) · COCO panoptic val2017PQ: 43.7 -> 49.3Improved
mix attention and FPN layers (3l3r) instead of FPN only (0l6r) (ablation)HyPiDecoder Mask DINO R50 (12 ep) · COCO panoptic val2017PQ: 48.7 -> 49.3Improved

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