TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/No time to train! Training-Free Reference-Based Instance S...

No time to train! Training-Free Reference-Based Instance Segmentation

Miguel Espinosa, Chenhongyi Yang, Linus Ericsson, Steven McDonagh, Elliot J. Crowley

2025-07-03Few-Shot Object DetectionSegmentationSemantic SegmentationInstance SegmentationCross-Domain Few-Shot Object DetectionImage Segmentation
PaperPDFCodeCode(official)

Abstract

The performance of image segmentation models has historically been constrained by the high cost of collecting large-scale annotated data. The Segment Anything Model (SAM) alleviates this original problem through a promptable, semantics-agnostic, segmentation paradigm and yet still requires manual visual-prompts or complex domain-dependent prompt-generation rules to process a new image. Towards reducing this new burden, our work investigates the task of object segmentation when provided with, alternatively, only a small set of reference images. Our key insight is to leverage strong semantic priors, as learned by foundation models, to identify corresponding regions between a reference and a target image. We find that correspondences enable automatic generation of instance-level segmentation masks for downstream tasks and instantiate our ideas via a multi-stage, training-free method incorporating (1) memory bank construction; (2) representation aggregation and (3) semantic-aware feature matching. Our experiments show significant improvements on segmentation metrics, leading to state-of-the-art performance on COCO FSOD (36.8% nAP), PASCAL VOC Few-Shot (71.2% nAP50) and outperforming existing training-free approaches on the Cross-Domain FSOD benchmark (22.4% nAP).

Results

TaskDatasetMetricValueModel
Object DetectionMS-COCO (1-shot)AP26.5Training-free
Object DetectionMS-COCO (30-shot)AP36.8Training-free
Object DetectionMS-COCO (10-shot)AP36.6Training-free
Object DetectionArtaxor mAP35Training-free(w/o FT)
Object DetectionNEU-DETmAP5.5Training-free(w/o FT)
Object DetectionDIORmAP16.4Training-free(w/o FT)
Object DetectionClipark1k mAP25.9Training-free(w/o FT)
Object DetectionDeepFishmAP29.6Training-free(w/o FT)
Object DetectionUODDmAP16Training-free(w/o FT)
3DMS-COCO (1-shot)AP26.5Training-free
3DMS-COCO (30-shot)AP36.8Training-free
3DMS-COCO (10-shot)AP36.6Training-free
3DArtaxor mAP35Training-free(w/o FT)
3DNEU-DETmAP5.5Training-free(w/o FT)
3DDIORmAP16.4Training-free(w/o FT)
3DClipark1k mAP25.9Training-free(w/o FT)
3DDeepFishmAP29.6Training-free(w/o FT)
3DUODDmAP16Training-free(w/o FT)
Few-Shot Object DetectionMS-COCO (1-shot)AP26.5Training-free
Few-Shot Object DetectionMS-COCO (30-shot)AP36.8Training-free
Few-Shot Object DetectionMS-COCO (10-shot)AP36.6Training-free
Few-Shot Object DetectionArtaxor mAP35Training-free(w/o FT)
Few-Shot Object DetectionNEU-DETmAP5.5Training-free(w/o FT)
Few-Shot Object DetectionDIORmAP16.4Training-free(w/o FT)
Few-Shot Object DetectionClipark1k mAP25.9Training-free(w/o FT)
Few-Shot Object DetectionDeepFishmAP29.6Training-free(w/o FT)
Few-Shot Object DetectionUODDmAP16Training-free(w/o FT)
2D ClassificationMS-COCO (1-shot)AP26.5Training-free
2D ClassificationMS-COCO (30-shot)AP36.8Training-free
2D ClassificationMS-COCO (10-shot)AP36.6Training-free
2D ClassificationArtaxor mAP35Training-free(w/o FT)
2D ClassificationNEU-DETmAP5.5Training-free(w/o FT)
2D ClassificationDIORmAP16.4Training-free(w/o FT)
2D ClassificationClipark1k mAP25.9Training-free(w/o FT)
2D ClassificationDeepFishmAP29.6Training-free(w/o FT)
2D ClassificationUODDmAP16Training-free(w/o FT)
2D Object DetectionMS-COCO (1-shot)AP26.5Training-free
2D Object DetectionMS-COCO (30-shot)AP36.8Training-free
2D Object DetectionMS-COCO (10-shot)AP36.6Training-free
2D Object DetectionArtaxor mAP35Training-free(w/o FT)
2D Object DetectionNEU-DETmAP5.5Training-free(w/o FT)
2D Object DetectionDIORmAP16.4Training-free(w/o FT)
2D Object DetectionClipark1k mAP25.9Training-free(w/o FT)
2D Object DetectionDeepFishmAP29.6Training-free(w/o FT)
2D Object DetectionUODDmAP16Training-free(w/o FT)
16kMS-COCO (1-shot)AP26.5Training-free
16kMS-COCO (30-shot)AP36.8Training-free
16kMS-COCO (10-shot)AP36.6Training-free
16kArtaxor mAP35Training-free(w/o FT)
16kNEU-DETmAP5.5Training-free(w/o FT)
16kDIORmAP16.4Training-free(w/o FT)
16kClipark1k mAP25.9Training-free(w/o FT)
16kDeepFishmAP29.6Training-free(w/o FT)
16kUODDmAP16Training-free(w/o FT)

Related Papers

SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction2025-07-21Deep Learning-Based Fetal Lung Segmentation from Diffusion-weighted MRI Images and Lung Maturity Evaluation for Fetal Growth Restriction2025-07-17DiffOSeg: Omni Medical Image Segmentation via Multi-Expert Collaboration Diffusion Model2025-07-17From Variability To Accuracy: Conditional Bernoulli Diffusion Models with Consensus-Driven Correction for Thin Structure Segmentation2025-07-17Unleashing Vision Foundation Models for Coronary Artery Segmentation: Parallel ViT-CNN Encoding and Variational Fusion2025-07-17SCORE: Scene Context Matters in Open-Vocabulary Remote Sensing Instance Segmentation2025-07-17Unified Medical Image Segmentation with State Space Modeling Snake2025-07-17A Privacy-Preserving Semantic-Segmentation Method Using Domain-Adaptation Technique2025-07-17