Accurately segmenting objects without any manual annotations remains one of the core challenges in computer vision. In this work, we introduce Selfment, a fully self-supervised framework that segments foreground objects directly from raw images without human labels, pretrained segmentation models, or any post-processing. Selfment first constructs patch-level affinity graphs from self-supervised features and applies NCut to obtain an initial coarse foreground--background separation. We then introduce Iterative Patch Optimization (IPO), a feature-space refinement procedure that progressively enforces spatial coherence and semantic consistency through iterative patch clustering. The refined masks are subsequently used as supervisory signals to train a lightweight segmentation head with contrastive and region-consistency objectives, allowing the model to learn stable and transferable object representations. Despite its simplicity and complete absence of manual supervision, Selfment sets new state-of-the-art (SoTA) results across multiple benchmarks. It achieves substantial improvements on Fmax over previous unsupervised saliency detection methods on ECSSD (+4.0%), HKUIS (+4.6%), and PASCAL-S (+5.7%). Moreover, without any additional fine-tuning, Selfment demonstrates remarkable zero-shot generalization to camouflaged object detection tasks (e.g., 0.910Sm on CHAMELEON and 0.792Fβω on CAMO), outperforming all existing unsupervised approaches and even rivaling the SoTA fully supervised methods.
Figures & tables
Figure 1: We propose Selfment , a fully self-supervised framework for foreground segmentation that generates highly detailed and accurate saliency maps without any human-annotated labels or post-processing.
Figure 2: An overview of Selfment. The input image is first encoded by a self-supervised backbone to produce dense patch features. These features define a patch-level affinity graph, from which we derive an initial foreground-background split using the second-smallest eigenvector of the NCut. We then apply Iterative Patch Optimization to improve spatial coherence and semantic consistency. The refined masks then serve as supervisory signals for training a lightweight segmentation head.
ECSSD [ 37 ]
DUTS [ 45 ]
HKUIS [ 21 ]
PASCAL-S [ 23 ]
Fmax
IoU
Acc.
Fmax
IoU
Acc.
Fmax
IoU
Acc.
Fmax
IoU
Acc.
HS [ 47 ]
67.3
50.8
84.7
50.4
36.9
82.6
-
-
-
-
-
-
wCtr [ 54 ]
68.4
51.7
86.2
52.2
39.2
83.5
-
-
-
-
-
-
WSC [ 22 ]
68.3
49.8
85.2
52.8
38.4
86.2
-
-
-
-
-
-
DeepUSPS [ 27 ]
58.4
44.0
79.5
42.5
30.5
77.3
-
-
-
-
-
-
BigBiGAN [ 44 ]
78.2
67.2
89.9
60.8
49.8
87.8
-
-
-
-
-
-
Table 1: Comparison of unsupervised saliency detection methods. We evaluate Selfment against state-of-the-art unsupervised methods on ECSSD [ 37 ] , DUTS [ 45 ] , HKUIS [ 21 ] , and PASCAL-S [ 23 ] . Best results are bolded .
Figure 3: Comparison with previous state-of-the-art methods on the unsupervised saliency detection task. All methods are evaluated without any post-processing at an inference resolution of 1280×1280 .
Figure 4: Comparison with previous state-of-the-art on the camouflaged object detection tasks.
CHAMELEON [ 42 ]
CAMO [ 19 ]
COD10K [ 9 ]
NC4K [ 26 ]
Sm↑
Fβω↑
Eξ↑
MAE
Sm↑
Fβω↑
Eξ↑
MAE
Sm↑
Fβω↑
Eξ↑
MAE
Sm↑
Fβω↑
Eξ↑
MAE
Fully-Supervised Methods
BGNet [ 43 ]
.901
.851
.954
.027
.812
.749
.870
.073
.831
.722
.901
.033
.851
.788
.907
.044
SINetv2 [ 8 ]
.888
.816
.961
.030
.820
.743
.882
.070
.815
.680
.887
.037
.847
.770
.903
.048
ZoomNet [ 29 ]
.902
.845
.958
.023
.820
.752
.878
.066
.838
.729
.888
.029
.853
.784
.896
.043
FSPNet [ 15 ]
.908
.851
.965
.023
.856
.799
.899
.050
.851
.735
.895
.026
.879
.816
.915
.035
Table 2: Comparison of camouflaged object detection methods. We evaluate Selfment against state-of-the-art methods on CHAMELEON [ 42 ] , CAMO [ 19 ] , COD10K [ 9 ] , and NC4K [ 26 ] . Best results are bolded .
Figure 5: Comparison of segmentation performance among TokenCut [ 46 ] , SelfMask [ 38 ] , FOUND [ 40 ] , and Selfment using DINO-Base, DINOv3-Huge+, and DINOv3-7B as backbones. Metrics are reported on the ECSSD [ 37 ] dataset.
Figure 6: Ablation study on the main components of the Selfment pipeline. From left to right, we progressively add NCut, Iterative Patch Optimization (IPO), and self-supervised learning (SSL).
Configuration
BCE
Dice
Con.
Fmax
IoU
Acc
NCut (baseline)
74.7
63.9
86.2
+ IPO
79.5
73.2
87.8
Self-supervised training
✓
88.3
80.4
94.4
+ Dice loss
✓
✓
88.9
81.3
94.7
+ Contrastive loss
✓
✓
88.9
81.4
94.7
+ Dice & Contrastive loss
✓
✓
✓
89.1
81.5
94.8
Table 3: Ablation study of individual components. Each module is added cumulatively to the previous configuration. BCE: binary cross-entropy loss; Dice: soft Dice loss; Con.: contrastive similarity loss.
Figure 7: An illustration showing the progressive refinement of the mask across iterations of the IPO procedure.
Fmax
IoU
Acc
<CLS>
29.2
20.5
54.0
K-Means
59.2
55.1
71.7
NCut
74.7
63.9
86.2
Table 4: Comparison of initial bipartition methods on segmentation performance. Metrics reported on the ECSSD dataset.
Figure 8: Qualitative comparison of how Selfment and TokenCut behave as input resolution increases.
Figure 9: Failure cases on unsupervised saliency detection.
The Segment Anything Model (SAM) relies heavily on massive manual annotations, creating a fundamental bottleneck for model scaling. While unsupervised methods attempt to learn object concepts from motion, they typically overfit to moving entities, lacking both multi-granularity understanding and the ability to generalize to static objects. To overcome this, we introduce Motion-Grounded Segment Anything (MoSA), a highly scalable unsupervised framework that learns a transferable objectness prior from unlabeled videos. MoSA operates in three progressive stages: (1) automatically generating multi-granularity motion pseudo-labels from large-scale video data; (2) training a Perceptual Grouping Model (PGM) via contrastive learning to internalize a generalized, appearance-driven concept of objects; and (3) transferring this learned prior into a prompt-guided architecture for segment-anything-style inference on images. Extensive zero-shot evaluations across seven challenging benchmarks (e.g., COCO and ADE20K) demonstrate that MoSA significantly outperforms existing unsupervised methods. Notably, despite using zero manual annotations, MoSA achieves segmentation performance comparable to the fully supervised SAM. Our findings reveal that harnessing large-scale unlabeled motion is a feasible and highly scalable alternative to annotation-driven segment-anything pipelines.
Weijian Jian, Xiaoyue Zhang, Bin Xiao +5
AI Research · Independent Researcher · University of Ottawa +1
Current zero-shot Camouflaged Object Segmentation methods typically employ a two-stage pipeline (discover-then-segment): using MLLMs to obtain visual prompts, followed by SAM segmentation. However, relying solely on MLLMs for camouflaged object discovery often leads to inaccurate localization, false positives, and missed detections. To address these issues, we propose the Discover-Segment-Select (DSS) mechanism, a progressive framework designed to refine segmentation step by step. The proposed method contains a Feature-coherent Object Discovery (FOD) module that leverages visual features to generate diverse object proposals, a segmentation module that refines these proposals through SAM segmentation, and a Semantic-driven Mask Selection (SMS) module that employs MLLMs to evaluate and select the optimal segmentation mask from multiple candidates. Without requiring any training or supervision, DSS achieves state-of-the-art performance on multiple COS benchmarks, especially in multiple-instance scenes.
Yilong Yang, Jianxin Tian, Shengchuan Zhang +1
Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, 361005, P.R. China.
In-context segmentation (ICS) aims to segment arbitrary concepts, e.g., objects, parts, or personalized instances, given one annotated visual examples. Existing work relies on (i) fine-tuning vision foundation models (VFMs), which improves in-domain results but harms generalization, or (ii) combines multiple frozen VFMs, which preserves generalization but yields architectural complexity and fixed segmentation granularities. We revisit ICS from a minimalist perspective and ask: Can a single self-supervised backbone support both semantic matching and segmentation, without any supervision or auxiliary models? We show that scaled-up dense self-supervised features from DINOv3 exhibit strong spatial structure and semantic correspondence. We introduce INSID3, a training-free approach that segments concepts at varying granularities only from frozen DINOv3 features, given an in-context example. INSID3 achieves state-of-the-art results across one-shot semantic, part, and personalized segmentation, outperforming previous work by +7.5 % mIoU, while using 3x fewer parameters and without any mask or category-level supervision. Code is available at https://github.com/visinf/INSID3 .
Claudia Cuttano, Gabriele Trivigno, Christoph Reich +3
Politecnico di Torino · TU Darmstadt · TU Munich +2