AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

MOCHA: Multi-modal Objects-aware Cross-arcHitecture Alignment

Elena Camuffo, Francesco Barbato, Mete Ozay, Simone Milani, Umberto Michieli

ECCV 2026 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary object detection · 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
add relational embedding loss L_emb (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 56.6 -> 59.1Improved
add relational embedding loss L_emb (ablation)MOCHA (AuXFT) · iCubWorld 1-shotretrieval accuracy: 59.8 -> 61.4Improved
use attention in the translator (ablation)MOCHA (AuXFT) · iCubWorld 5-shotretrieval accuracy: 73.2 -> 77.0Improved
keep final ffn layer in translator (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 55.7 -> 59.1Improved
use non-linear translator instead of linear projection (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 49.3 -> 59.1Improved
use non-linear translator instead of linear projection (ablation)MOCHA (AuXFT) · POD 1-shotmAP50-95: 29.4 -> 36.3Improved
use all pairs in relational loss instead of Top-K=32 (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 51.1 -> 59.1Improved
do not enable data augmentation during distillation (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 53.3 -> 59.1Improved
keep default learning rate instead of doubling encoder lr (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 53.6 -> 59.1Improved
use PCA dimension 512 instead of 1024 (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 53.8 -> 59.1Improved
use PCA dimension 512 instead of 256 (ablation)MOCHA (AuXFT) · PerSeg 1-shotmAP50-95: 56.4 -> 59.1Improved
pretrain student on COCO instead of OpenImages only (ablation)MOCHA · PerSeg 1-shotmAP50-95: 53.8 -> 55.4Improved

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