AI Paper Atlas

Research map / Detection and segmentation / CNN features

MOBIUS: Big-to-Mobile Universal Instance Segmentation via Multi-modal Bottleneck Fusion and Calibrated Decoder Pruning

Mattia Segu, Marta Tintore Gazulla, Yongqin Xian, Luc Van Gool, Federico Tombari

ICCV 2025 · Detection and segmentation › CNN features · approach: Boundary-aware instance segmentation · 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
replace dot product with scaled cosine similarity for language-guided classification (ablation)multi-modal DETR generalist (GLEE-style) · COCO-valconvergence: did not converge -> 45.9 AP_boxImproved
add scaled cosine only, no calibration, in joint single-stage training (ablation)MOBIUS-H-R50 · Joint training (all datasets)convergence: D.N.C.No clear effect
add language-guided uncertainty calibration loss on top of scaled cosine (ablation)MOBIUS-H-R50 · Joint training, single stageconvergence / AP_box: D.N.C. -> 50.0 AP_box COCO-valImproved
add calibration loss on COCO-only training (ablation)MOBIUS-H-R50 · COCO-valAP_box: 45.9 -> 46.5Improved
decode a single scale instead of multi-scale (ablation)MaskDINO pixel decoder · COCO-valAP_box: 49.2 -> 47.9Hurt
decode a single scale instead of multi-scale (ablation)MOBIUS bottleneck encoder · COCO-valAP_box: 47.9 -> 47.5No clear effect
use stride-16 bottleneck instead of stride-8 (ablation)MOBIUS bottleneck encoder · COCO-valAP_box: 47.5 -> 47.5 with 200 -> 91 pixel decoder GFLOPsNo clear effect
reduce MaskDINO pixel decoder to 1 layer (ablation)MaskDINO pixel decoder (single-scale) · COCO-valAP_box: 47.9 -> 43.4Hurt
swap MaskDINO pixel decoder for RT-DETR pixel decoder (ablation)GLEE (mobile MNv4-CM) · LVIS / ODinWAP_box: LVIS 28.9 -> 22.8; ODinW 37.0 -> 30.7Hurt
replace MaskDINO pixel decoder with bottleneck encoder (ablation)GLEE (mobile MNv4-CM) · ODinWAP_box: 37.0 -> 38.6Improved
replace MaskDINO pixel decoder with bottleneck encoderGLEE (mobile MNv4-CM) · Samsung S24 latencylatency (ms): 436.6 -> 127.1Improved
add sigmoidal-growth inference-time query pruning (ablation)MOBIUS transformer decoder · COCO-valAP_mask and decoder FLOPs: decoder FLOPs reduced by additional 50%, AP_mask 44.0Improved

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