Attention-disentangled Uniform Orthogonal Feature Space Optimization for Few-shot Object Detection

Fanman Meng; Heqian Qiu; Hongliang Li; Lanxiao Wang; Qingbo Wu; Taijin Zhao; Yu Dai

REVIEW 3 major objections 5 minor 46 references

Attention-disentangled Uniform Orthogonal Feature Space Optimization for Few-shot Object Detection

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Decoupling objectness (feature magnitude) from classification (feature angle) lets few-shot detectors transfer objectness knowledge from base to novel classes.

desk verdict A genuinely new low-shot FSOD pipeline with impressive 1-3 shot gains, but the ablation doesn't isolate the claimed mechanism and the 'significant' claim is too broad. read the letter →

arxiv 2506.22161 v1 pith:MZ2YFOUZ submitted 2025-06-27 cs.CV

Taijin Zhao , Heqian Qiu , Yu Dai , Lanxiao Wang , Fanman Meng , Qingbo Wu , Hongliang Li This is my paper · ORCID

classification cs.CV

keywords few-shotobjectdetectionorthogonalfeaturespaceobjectnesstransferdisentanglementbackgroundoptimizationtask-specificattentioncrop-pasteaugmentation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that few-shot object detection suffers because standard detectors entangle objectness ('is there an object at all?') with classification ('which class is it?') in one feature space, so the objectness criterion gets tuned to the abundant base classes and fails on novel classes that have only a few samples. It proposes a Uniform Orthogonal Feature Space in which the magnitude of a ROI feature predicts objectness and the angle of that feature against class prototypes predicts the class, making the objectness criterion class-agnostic and transferable from base to novel classes. The paper adds a hybrid background optimization: a set of 'pure background' images is constructed by segmenting out annotated objects and pasting them on neutral gray backgrounds, giving clean magnitude supervision, while hidden foreground instances in the original images are used to spread the angular distribution. A spatial attention module further separates class-agnostic features (objectness and box regression) from class-specific features (classification). The reported experiments on PASCAL VOC and MS COCO show the largest gains in the low-shot settings, supporting the claim that the disentanglement is what helps.

What carries the argument

The load-bearing mechanism is the Uniform Orthogonal Feature Space (UOFS): feature magnitude encodes objectness and feature angle encodes classification. The magnitude branch uses $p_{obj} = -|W_{obj}|\,\|f\| + b$, so larger feature vectors mean lower objectness; the angle branch uses cosine similarity between the normalized ROI feature $f/\|f\|$ and learnable class prototypes $W_{cls}$, with a temperature $\tau=20$. Two supporting mechanisms make this space trainable: Hybrid Background Optimization (HBO), which builds a pure-background base set by SAM-2-based crop-paste of annotated instances onto normalized gray backgrounds and jointly trains on both that set and the original base set with learnable unknown-class prototypes $W_{unk}$; and the Spatial-wise Attention Disentanglement and Association (SADA) module, which applies spatial attention masks to produce class-specific features for classification and class-agnostic features for objectness and box regression.

What would settle it

Train the detector on base classes only, then test on held-out novel-class images without any novel fine-tuning: if the magnitude-based objectness score does not rank true novel instances above background, the claimed class-agnostic objectness transfer is not occurring. A second check is to build the pure-background set with random crops instead of SAM-2 masks (or with masks from a different segmentation model) and compare low-shot nAP; if performance is unchanged, the purity of the background set is not the cause of the reported gains.

Watch

Extended reading notes

Core claim

The paper's central claim is that the entanglement of objectness and classification inside a shared classifier is the main bottleneck for few-shot object detection, and that decoupling the two in an orthogonal feature space removes that bottleneck. Concretely, each ROI feature $f$ is decomposed so that $\|f\|$ alone drives the objectness probability $p_{obj} = -|W_{obj}| \|f\| + b$ and the normalized direction $f/\|f\|$ drives class probabilities via cosine similarity with class prototypes; the final foreground score is $p_{obj} \cdot p_{cls}$ and the background score is $1 - p_{obj}$. Because objectness is computed without any class information, it can be learned from abundant base classes and then applied directly to novel classes, avoiding the bias that comes from tuning objectness criteria on a handful of unrepresentative novel instances. The paper also claims that this transfer requires two further ingredients: a pure-background base set that removes unlabeled foreground instances from the objectness supervision, and the use of those same unlabeled instances, plus learnable unknown-class prototypes, to keep the angular layout uniform so that novel classes do not collapse into base-class clusters. On PASCAL VOC and MS COCO the method reports consistent low-shot improvements over its DeFRCN baseline, with the caveat that high-shot gains are smaller and sometimes negative.

Load-bearing premise

The method assumes the pure-background base set, built by segmenting annotated objects and pasting them on normalized gray backgrounds, contains no leftover foreground objects and therefore gives an unbiased signal for what counts as an object.

Editorial extensions

If this is right

At 1-shot on PASCAL VOC Split1, the method reports a single-run nAP50 of 64.3, compared to 57.0 for the DeFRCN baseline; the reported gains are largest at 1/2/3-shot settings.
Since objectness is a class-agnostic magnitude score, novel-class fine-tuning only needs to adjust angular prototypes instead of relearning what a foreground object is.
The learnable unknown-class prototypes and pure-background set raise the Silhouette Coefficient on all classes from 0.1018 to 0.1519 relative to plain orthogonal features, which the paper interprets as a more uniform angular distribution ready for novel classes.
The method's weaker performance in Split2 high-shot settings is attributed to small instance sizes such as bottle, indicating a residual size sensitivity in the magnitude criterion.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension the paper does not test is freezing the objectness branch after base training and only updating class prototypes during novel fine-tuning; if the transfer claim is right, this should lose little accuracy and test the mechanism directly.
The dependency on SAM-2 means the method's gains could degrade if the segmentation model leaves unlabeled foreground objects in the 'pure' backgrounds; measuring the surviving foreground rate in the pure-background set would quantify this risk.
The same magnitude/angle split could be applied to open-vocabulary or incremental detection, where objectness must persist while class vocabularies grow; the paper only evaluates the closed-set FSOD setting.
The hyperparameter search over the number of unknown prototypes (best at 5) suggests the method's benefit may depend on matching the capacity of the unknown-class buffer to the actual number of unseen classes in the target set.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Here's my take. The paper brings something new to FSOD: it takes the orthogonal feature space idea from OrthogonalDet and adapts it to few-shot detection, using a SAM-2-constructed pure background set to give clean objectness supervision and learnable unknown-class prototypes to absorb unlabeled instances. That combination is not routine, and the low-shot numbers are genuinely impressive—MS COCO 1-shot nAP 10.3 vs 6.5 for DeFRCN, and PASCAL Split1 1-shot 64.3 vs 57.0. The gains are consistent across both benchmarks at 1-3 shots, and the authors even offer a plausible explanation for why high-shot performance sometimes drops (tiny instances in Split2, positive/negative imbalance).

The soft spots are real, though. The biggest one is the ablation in Table IV. UOFS is compared to OFS, but UOFS adds the PB base set, unknown prototypes, HBO, and SADA all at once. The jump from 38.5 to 47.3 could come from the synthetic background data acting as augmentation, not from the magnitude/angle disentanglement itself. The paper never runs the control where the PB set/HBO is applied to a Euclidean or cosine head. That leaves the central causal claim underdetermined. The stress-test note says this; I agree with it.

Second, the abstract says 'significantly outperforms,' but the tables show high-shot rows where it loses (VOC Split2 10-shot 51.0 vs 52.9, COCO 30-shot 21.0 vs 21.0). The claim is only true at very low shot counts, and even there we don't get error bars on the multiple-run averages, so we can't judge statistical significance.

Third, several hyperparameters (alpha, number of unknown prototypes, temperature, negative placement, background type) are tuned on the test benchmarks without an obvious held-out validation split. That's common in this literature, but it still weakens the comparative claims.

The SAM-2 purity issue is acknowledged but not stress-tested. If the pure background set still contains unlabeled instances, the objectness supervision is biased. That's a minor concern because the method works anyway, but it's worth probing.

Overall: this is a serious paper with a worthwhile idea and strong low-shot results. It deserves a proper peer review, but it needs major revision to support the mechanistic claim and to temper the significance statement. I'd want to see the missing control, error bars, and a validation split for hyperparameters before believing the disentanglement is the real driver. As is, it's a conditional accept at best.

Referee Report

3 major / 5 minor

Summary. The paper proposes a Uniform Orthogonal Feature Space (UOFS) framework for few-shot object detection. The key idea is to decouple objectness and classification by letting feature magnitude encode objectness and feature angle encode class identity, in contrast to the entangled Euclidean or cosine spaces used in typical Faster R-CNN-based FSOD methods. To support this disentanglement, the authors construct a 'pure background' base set by segmenting base-class instances with SAM-2 and pasting them onto normalized gray backgrounds, and they introduce learnable unknown-class prototypes and a hybrid background optimization (HBO) loss. A Spatial-wise Attention Disentanglement and Association (SADA) module is added to separate class-agnostic and class-specific ROI features. Experiments on PASCAL VOC and MS COCO report consistent low-shot gains over the DeFRCN baseline and several prior methods, but the paper also shows losses or ties in some high-shot settings.

Significance. If the mechanism were convincingly isolated, the idea of transferring class-agnostic objectness via magnitude/angle separation would be a valuable contribution to few-shot object detection, and the reported low-shot improvements are consistent across two benchmarks. The paper also provides useful empirical comparisons of background designs, unknown-prototype counts, and negative-sample placement. However, the current experimental design does not isolate the disentanglement mechanism from the data-augmentation effects of the PB base set, and the abstract's unqualified claim of significant overall outperformance is contradicted by some high-shot rows. The method is therefore promising but the central causal claim needs stronger experimental support.

major comments (3)

[V-D, Table IV] The central claim that magnitude/angle disentanglement is the cause of the improvement rests on Table IV, where UOFS outperforms OFS by 8.8 points at 1-shot (47.3 vs. 38.5). However, UOFS is not OFS plus orthogonality alone; it also adds the PB base set (Sec. IV-C), the unknown-class prototypes and HBO (Sec. IV-D), and SADA (Sec. IV-E). Because no control applies the PB set and SADA to the cosine or Euclidean heads, the observed gain could be driven by the synthetic background data and loss reweighting rather than by the disentangled objectness mechanism. Please add ablations such as Euclidean+PB-set, cosine+PB-set, and OFS+PB-set without the unknown prototypes to isolate the contribution of the orthogonal decomposition.
[Abstract, V-C (Tables I and II)] The abstract's claim that the method 'significantly outperforms existing approaches based on entangled feature spaces' is not supported in high-shot regimes: on PASCAL VOC Split2 10-shot the method is below DeFRCN (51.2 vs. 52.9 single-run; 51.0 vs. 52.9 multiple-run), and on MS COCO 30-shot it ties or falls below DeFRCN (21.0 vs. 21.0 nAP multiple-run; 21.8 vs. 22.5 single-run). In addition, the multiple-run tables report only averages, without standard deviations, so the word 'significantly' is not statistically justified. The claims should be restricted to low-shot settings and accompanied by variance estimates.
[IV-C, IV-D] The method assumes that the PB base set is free of unlabeled foreground instances and therefore provides unbiased magnitude supervision. The authors acknowledge that SAM-2 segmentation is imperfect (Sec. IV-D) but never measure how many unlabeled instances remain in the constructed PB set or how their presence affects the objectness criterion. This is load-bearing for the objectness-transfer story. Please report a purity statistic for the PB set (e.g., fraction of crops that still contain non-base foreground pixels by human or oracle inspection) and test sensitivity to the SAM-2 threshold or to using oracle masks.

minor comments (5)

[IV-D, Eq. (9)] Equation (9) uses 'pcls[:, : Nc]' but this should be 'pcls-unk[:, : Nc]' to match the definition in the preceding lines; the current notation is confusing because pcls in Eq. (7) has only Nc entries.
[Table VI caption] The caption 'IMPACT OF THE NUMBER OF UNKNOWN CLASS PROTOTYPE' refers to 'Num. learn bg', but the table rows are numbers of unknown prototypes; 'learn bg' should be 'learnable unknown prototypes' for clarity.
[Table IX] The 'w/o' row in Table IX is ambiguous; since the table reports the effectiveness of SADA, the row should be labeled 'w/o SADA' to make clear that it is the UOFS baseline without the module.
[Fig. 4] The text for the negative-sample placement strategy refers to panels (a), (b), and (c), but the caption and the figure itself do not label which panel corresponds to the 'inner' or 'outer' placement; please add labels or describe all panels.
[V-B] The implementation details give fine-tuning iterations for MS COCO but not for PASCAL VOC; please report the corresponding iteration counts for VOC to allow replication.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central FSOD claim is evaluated against external baselines, and self-citations are not load-bearing.

full rationale

The paper's central empirical claim, that UOFS improves few-shot object detection, is supported by comparisons against external methods (TFA, DeFRCN, FSCE, ECEA, etc.) on PASCAL VOC and MS COCO. The orthogonal feature space idea is explicitly borrowed from external prior work (OrthogonalDet [13], Norm-Aware Embedding [14], LDET [27]) rather than being asserted only by self-citation. The only self-citations, [11] and [35], appear in related-work descriptions and attention-module motivation, and they do not carry the derivation of the main results. The unknown-class prototype mechanism (Eqs. 8-9) is a trained model component, not a prediction that reduces to its own definition: it is neither fitted to novel-class performance nor used to define the evaluation metric. The ablation in Table IV compares UOFS against OFS while also introducing the PB base set, HBO, and SADA, so the improvement is not fully isolated causally; however, this is a confounding/missing-control issue, not circularity, because no equation makes the reported gain equal to its input by construction. The acknowledged limitations (high-shot underperformance on VOC Split 2 and COCO 30-shot, SAM-2 segmentation imperfections, and residual foregrounds in BG-20K) weaken the strength of the empirical claims but do not make the derivation circular. Overall, no load-bearing step reduces to its own inputs or to a self-citation chain, so the circularity score is 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The central claim rests on several design postulates rather than on a closed-form derivation. The free parameters (alpha, Nu, tau, background type, negative placement) are tuning choices made on the benchmark test sets. The entity Wunk (unknown-class prototypes) is internal to the training loss and has no external falsifiable handle. The key domain assumptions (G-FSOD setting, unlabeled instances in base images) are standard in the field and stated explicitly.

free parameters (5)

alpha (loss weight) = 0.5
Controls the balance between base set and pure background base set losses in Eq. (4); empirically set without an ablation study.
Nu (number of unknown class prototypes) = 5
Number of learnable prototypes Wunk for unknown classes; chosen by test-set ablation (Table VI), although Nu=3 gives higher 1-shot score (47.8 vs 47.3).
tau (classification temperature) = 20
Softmax temperature in Eq. (6) and (8); adopted from TFA [6] rather than tuned in this paper.
negative sample placement = outer hypersphere
Design choice of placing negative samples on outer vs inner hypersphere regions, selected by test-set comparison (Table VII).
background type for PB base set = normalized gray
Chosen from three alternatives by best test-set performance (Table III); the rationale that natural backgrounds still contain foregrounds is inferred after observing results.

assumptions (5)

ad hoc to paper The feature space can be decomposed such that magnitude encodes objectness and angle encodes classification through the loss in Eq. (2)-(7).
This is a design choice enforced by the training objective, not a property derived from the data.
ad hoc to paper SAM-2 segmentation and crop-paste onto normalized gray backgrounds produce a pure background base set without unlabeled foreground instances.
The central objectness supervision assumes the PB base set is free of hidden novel-class instances; the paper acknowledges imperfect segmentation in Section IV-D but does not analyze its impact.
ad hoc to paper Unlabeled foreground instances in the base set are sufficiently well captured by Nu=5 learnable unknown-class prototypes during joint training.
The angular uniformity claim depends on these prototypes absorbing hard negatives; no external evidence links the prototypes to actual novel classes.
domain assumption The G-FSOD task setting and Faster R-CNN architecture are suitable backbones for the proposed disentanglement.
The method is built on DeFRCN and Faster R-CNN following prior FSOD work; the orthogonality design may not transfer to other detectors.
domain assumption Base and novel classes are mutually exclusive, and base set images may contain unlabeled instances of novel classes.
This is the standard FSOD assumption, stated in Section III-A, and is a precondition for the label-missing problem the paper targets.

invented entities (2)

Unknown class prototypes Wunk (Nu=5)
purpose: Represent unlabeled foreground instances (potential novel classes) in the angular feature space, preventing them from collapsing into base class clusters.
Introduced specifically for HBO; no falsifiable prediction outside the paper's own benchmarks, and the number of prototypes is a tuned hyperparameter.
Pure background base set (PB base set)
purpose: Provide unbiased magnitude-based objectness supervision by removing unlabeled foreground instances through crop-paste of base class instances.
A constructed dataset artifact; its effectiveness is only demonstrated on the target benchmarks and it is not released.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Attention-disentangled Uniform Orthogonal Feature Space Optimization for Few-shot Object Detection." pith.science (2026). https://pith.science/paper/MZ2YFOUZ

@misc{pith2026250622161,
  author       = {Pith},
  title        = {Pith review of: Attention-disentangled Uniform Orthogonal Feature Space Optimization for Few-shot Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MZ2YFOUZ}},
  note         = {Machine review of arXiv:2506.22161}
}

read the original abstract

Few-shot object detection (FSOD) aims to detect objects with limited samples for novel classes, while relying on abundant data for base classes. Existing FSOD approaches, predominantly built on the Faster R-CNN detector, entangle objectness recognition and foreground classification within shared feature spaces. This paradigm inherently establishes class-specific objectness criteria and suffers from unrepresentative novel class samples. To resolve this limitation, we propose a Uniform Orthogonal Feature Space (UOFS) optimization framework. First, UOFS decouples the feature space into two orthogonal components, where magnitude encodes objectness and angle encodes classification. This decoupling enables transferring class-agnostic objectness knowledge from base classes to novel classes. Moreover, implementing the disentanglement requires careful attention to two challenges: (1) Base set images contain unlabeled foreground instances, causing confusion between potential novel class instances and backgrounds. (2) Angular optimization depends exclusively on base class foreground instances, inducing overfitting of angular distributions to base classes. To address these challenges, we propose a Hybrid Background Optimization (HBO) strategy: (1) Constructing a pure background base set by removing unlabeled instances in original images to provide unbiased magnitude-based objectness supervision. (2) Incorporating unlabeled foreground instances in the original base set into angular optimization to enhance distribution uniformity. Additionally, we propose a Spatial-wise Attention Disentanglement and Association (SADA) module to address task conflicts between class-agnostic and class-specific tasks. Experiments demonstrate that our method significantly outperforms existing approaches based on entangled feature spaces.

Figures

Figures reproduced from arXiv: 2506.22161 by the authors.

**Figure 2.** The framework of our proposed method. Images from the base set and the pure background base set are sent into the detector simultaneously. The pure [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** The pipeline of constructing the pure background base set. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Illustration of the impact of hyperspherical regions. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: Visualization results of disentangled class-specific attention map for [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 42 canonical work pages

[1]

Yolo9000: better, faster, stronger,

J. Redmon and A. Farhadi, “Yolo9000: better, faster, stronger,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 7263–7271

work page 2017
[2]

Crossdet++: Growing crossline representation for object detection,

H. Qiu, H. Li, Q. Wu, J. Cui, Z. Song, L. Wang, and M. Zhang, “Crossdet++: Growing crossline representation for object detection,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 33, no. 3, pp. 1093–1108, 2022

work page 2022
[3]

Fast R-CNN,

R. B. Girshick, “Fast R-CNN,” in Proceedings of the IEEE International Conference on Computer Vision , 2015, pp. 1440–1448

work page 2015
[4]

Faster r-cnn: Towards real-time object detection with region proposal networks,

S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1137– 1149, 2016

work page 2016
[5]

Cascade r-cnn: Delving into high quality object detection,

Z. Cai and N. Vasconcelos, “Cascade r-cnn: Delving into high quality object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2018, pp. 6154–6162

work page 2018
[6]

Frustratingly simple few-shot object detection,

X. Wang, T. Huang, J. Gonzalez, T. Darrell, and F. Yu, “Frustratingly simple few-shot object detection,” in Proceedings of the International Conference on Machine Learning . PMLR, 2020, pp. 9919–9928

work page 2020
[7]

Defrcn: Decoupled faster r-cnn for few-shot object detection,

L. Qiao, Y . Zhao, Z. Li, X. Qiu, J. Wu, and C. Zhang, “Defrcn: Decoupled faster r-cnn for few-shot object detection,” in Proceedings of the IEEE International Conference on Computer Vision , 2021, pp. 8681–8690

work page 2021
[8]

Few-shot object detection via association and discrimination,

Y . Cao, J. Wang, Y . Jin, T. Wu, K. Chen, Z. Liu, and D. Lin, “Few-shot object detection via association and discrimination,” Advances in Neural Information Processing Systems , vol. 34, pp. 16 570–16 581, 2021

work page 2021

Show all 46 references

[9]

Label, verify, correct: A simple few shot object detection method,

P. Kaul, W. Xie, and A. Zisserman, “Label, verify, correct: A simple few shot object detection method,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2022, pp. 14 237–14 247

2022
[10]

Semi-supervised few-shot object detection via adaptive pseudo labeling,

Y . Tang, Z. Cao, Y . Yang, J. Liu, and J. Yu, “Semi-supervised few-shot object detection via adaptive pseudo labeling,” IEEE Transactions on Circuits and Systems for Video Technology , 2023

2023
[11]

Vlm-guided explicit-implicit complementary novel class se- mantic learning for few-shot object detection,

T. Zhao, H. Qiu, Y . Dai, L. Wang, H. Mei, F. Meng, Q. Wu, and H. Li, “Vlm-guided explicit-implicit complementary novel class se- mantic learning for few-shot object detection,” Expert Systems with Applications, vol. 256, p. 124926, 2024

2024
[12]

Ecea: Extensible co-existing attention for few-shot object detection,

Z. Xin, T. Wu, S. Chen, Y . Zou, L. Shao, and X. You, “Ecea: Extensible co-existing attention for few-shot object detection,” IEEE Transactions on Image Processing , 2024

2024
[13]

Exploring orthogonality in open world object detection,

Z. Sun, J. Li, and Y . Mu, “Exploring orthogonality in open world object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2024, pp. 17 302–17 312

2024
[14]

Norm-aware embedding for efficient person search,

D. Chen, S. Zhang, J. Yang, and B. Schiele, “Norm-aware embedding for efficient person search,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2020, pp. 12 615–12 624

2020
[15]

Few-shot object detection via feature reweighting,

B. Kang, Z. Liu, X. Wang, F. Yu, J. Feng, and T. Darrell, “Few-shot object detection via feature reweighting,” in Proceedings of the IEEE International Conference on Computer Vision , 2019, pp. 8420–8429

2019
[16]

Beyond max-margin: Class margin equilibrium for few-shot object detection,

B. Li, B. Yang, C. Liu, F. Liu, R. Ji, and Q. Ye, “Beyond max-margin: Class margin equilibrium for few-shot object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2021, pp. 7363–7372

2021
[17]

Few-shot object detection: Research advances and challenges,

Z. Xin, S. Chen, T. Wu, Y . Shao, W. Ding, and X. You, “Few-shot object detection: Research advances and challenges,” Information Fusion , p. 102307, 2024

2024
[18]

Adaptive subspaces for few-shot learning,

C. Simon, P. Koniusz, R. Nock, and M. Harandi, “Adaptive subspaces for few-shot learning,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2020, pp. 4136–4145

2020
[19]

Constrained few-shot class-incremental learning,

M. Hersche, G. Karunaratne, G. Cherubini, L. Benini, A. Sebastian, and A. Rahimi, “Constrained few-shot class-incremental learning,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2022, pp. 9057–9067

2022
[20]

Orthogonal projection loss,

K. Ranasinghe, M. Naseer, M. Hayat, S. Khan, and F. S. Khan, “Orthogonal projection loss,” in Proceedings of the IEEE International Conference on Computer Vision , 2021, pp. 12 333–12 343

2021
[21]

Learning orthogonal prototypes for generalized few-shot semantic segmentation,

S.-A. Liu, Y . Zhang, Z. Qiu, H. Xie, Y . Zhang, and T. Yao, “Learning orthogonal prototypes for generalized few-shot semantic segmentation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2023, pp. 11 319–11 328

2023
[22]

Generating features with increased crop- related diversity for few-shot object detection,

J. Xu, H. Le, and D. Samaras, “Generating features with increased crop- related diversity for few-shot object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 713–19 722

2023
[23]

Simple copy-paste is a strong data augmentation method for instance segmentation,

G. Ghiasi, Y . Cui, A. Srinivas, R. Qian, T.-Y . Lin, E. D. Cubuk, Q. V . Le, and B. Zoph, “Simple copy-paste is a strong data augmentation method for instance segmentation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2021, pp. 2918–2928

2021
[24]

X-paste: Revisiting scalable copy-paste for instance segmentation using clip and stablediffusion,

H. Zhao, D. Sheng, J. Bao, D. Chen, D. Chen, F. Wen, L. Yuan, C. Liu, W. Zhou, Q. Chu et al. , “X-paste: Revisiting scalable copy-paste for instance segmentation using clip and stablediffusion,” in Proceedings of the International Conference on Machine Learning . PMLR, 2023, p...

2023
[25]

Pin: Posi- tional insert unlocks object localisation abilities in vlms,

M. Dorkenwald, N. Barazani, C. G. Snoek, and Y . M. Asano, “Pin: Posi- tional insert unlocks object localisation abilities in vlms,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2024, pp. 13 548–13 558

2024
[26]

An effective crop-paste pipeline for few-shot object detection,

S. Lin, K. Wang, X. Zeng, and R. Zhao, “An effective crop-paste pipeline for few-shot object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2023, pp. 4820–4828

2023
[27]

Learning to detect every thing in an open world,

K. Saito, P. Hu, T. Darrell, and K. Saenko, “Learning to detect every thing in an open world,” in Proceedings of the European Conference on Computer Vision. Springer, 2022, pp. 268–284

2022
[28]

Fsce: Few-shot object detection via contrastive proposal encoding,

B. Sun, B. Li, S. Cai, Y . Yuan, and C. Zhang, “Fsce: Few-shot object detection via contrastive proposal encoding,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2021, pp. 7352–7362

2021
[29]

Orthogonal progressive network for few-shot object detection,

B. Wang and D. Yu, “Orthogonal progressive network for few-shot object detection,” Expert Systems with Applications , p. 125905, 2025

2025
[30]

Cat: Localization and identification cascade detection transformer for open- world object detection,

S. Ma, Y . Wang, Y . Wei, J. Fan, T. H. Li, H. Liu, and F. Lv, “Cat: Localization and identification cascade detection transformer for open- world object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 681–19 690

2023
[31]

Sam 2: Segment anything in images and videos,

N. Ravi, V . Gabeur, Y .-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. R ¨adle, C. Rolland, L. Gustafson, E. Mintun, J. Pan, K. V . Alwala, N. Carion, C.-Y . Wu, R. Girshick, P. Doll ´ar, and C. Feichtenhofer, “Sam 2: Segment anything in images and videos,” arXiv preprint arXiv:...

2024 arXiv
[32]

Bridging composite and real: towards end-to-end deep image matting,

J. Li, J. Zhang, S. J. Maybank, and D. Tao, “Bridging composite and real: towards end-to-end deep image matting,” International Journal of Computer Vision, vol. 130, no. 2, pp. 246–266, 2022

2022
[33]

Multi-faceted distillation of base-novel commonality for few-shot object detection,

S. Wu, W. Pei, D. Mei, F. Chen, J. Tian, and G. Lu, “Multi-faceted distillation of base-novel commonality for few-shot object detection,” in Proceedings of the European Conference on Computer Vision. Springer, 2022, pp. 578–594

2022
[34]

Few-shot object detection via variational feature aggregation,

J. Han, Y . Ren, J. Ding, K. Yan, and G.-S. Xia, “Few-shot object detection via variational feature aggregation,” inProceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 1, 2023, pp. 755–763

2023
[35]

Mcce-rec: Mllm- driven cross-modal contrastive entropy model for zero-shot referring expression comprehension,

H. Qiu, L. Wang, T. Zhao, F. Meng, Q. Wu, and H. Li, “Mcce-rec: Mllm- driven cross-modal contrastive entropy model for zero-shot referring expression comprehension,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

2024
[36]

Meta-detr: Image- level few-shot detection with inter-class correlation exploitation,

G. Zhang, Z. Luo, K. Cui, S. Lu, and E. P. Xing, “Meta-detr: Image- level few-shot detection with inter-class correlation exploitation,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, no. 11, pp. 12 832–12 843, 2022

2022
[37]

Fine-grained prototypes distillation for few-shot object detection,

Z. Wang, B. Yang, H. Yue, and Z. Ma, “Fine-grained prototypes distillation for few-shot object detection,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 6, 2024, pp. 5859– 5866

2024
[38]

Understanding negative proposals in generic few-shot object detection,

B. Yan, C. Lang, G. Cheng, and J. Han, “Understanding negative proposals in generic few-shot object detection,” IEEE Transactions on Circuits and Systems for Video Technology , 2024

2024
[39]

Fsna: Few-shot object detection via neighborhood information adaption and all attention,

J. Zhu, Q. Wang, X. Dong, W. Ruan, H. Chen, L. Lei, and G. Hao, “Fsna: Few-shot object detection via neighborhood information adaption and all attention,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

2024
[40]

Multi-view part-based few-shot object detection,

J. Ma and S. Bai, “Multi-view part-based few-shot object detection,” IEEE Transactions on Neural Networks and Learning Systems , 2024

2024
[41]

Text generation and multi-modal knowledge transfer for few- shot object detection,

Y . Du, F. Liu, L. Jiao, S. Li, Z. Hao, P. Li, J. Wang, H. Wang, and X. Liu, “Text generation and multi-modal knowledge transfer for few- shot object detection,” Pattern Recognition, p. 111283, 2025. JOURNAL OF LATEX CLASS FILES 11

2025
[42]

The pascal visual object classes (voc) challenge,

M. Everingham, L. Van Gool, C. K. Williams, J. Winn, and A. Zisser- man, “The pascal visual object classes (voc) challenge,” International Journal of Computer Vision , vol. 88, pp. 303–338, 2010

2010
[43]

Microsoft coco: Common objects in context,

T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Proceedings of the European Conference on Computer Vision. Springer, 2014, pp. 740–755

2014
[44]

Multi-scale positive sample re- finement for few-shot object detection,

J. Wu, S. Liu, D. Huang, and Y . Wang, “Multi-scale positive sample re- finement for few-shot object detection,” in Proceedings of the European Conference on Computer Vision , 2020, pp. 456–472

2020
[45]

Disentangle and remerge: interventional knowledge distillation for few-shot object detection from a conditional causal perspective,

J. Li, Y . Zhang, W. Qiang, L. Si, C. Jiao, X. Hu, C. Zheng, and F. Sun, “Disentangle and remerge: interventional knowledge distillation for few-shot object detection from a conditional causal perspective,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol...

2023
[46]

Deep residual learning for image recognition,

K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

2016

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Yolo9000: better, faster, stronger,

J. Redmon and A. Farhadi, “Yolo9000: better, faster, stronger,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 7263–7271

work page 2017

[2] [2]

Crossdet++: Growing crossline representation for object detection,

H. Qiu, H. Li, Q. Wu, J. Cui, Z. Song, L. Wang, and M. Zhang, “Crossdet++: Growing crossline representation for object detection,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 33, no. 3, pp. 1093–1108, 2022

work page 2022

[3] [3]

Fast R-CNN,

R. B. Girshick, “Fast R-CNN,” in Proceedings of the IEEE International Conference on Computer Vision , 2015, pp. 1440–1448

work page 2015

[4] [4]

Faster r-cnn: Towards real-time object detection with region proposal networks,

S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1137– 1149, 2016

work page 2016

[5] [5]

Cascade r-cnn: Delving into high quality object detection,

Z. Cai and N. Vasconcelos, “Cascade r-cnn: Delving into high quality object detection,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2018, pp. 6154–6162

work page 2018

[6] [6]

Frustratingly simple few-shot object detection,

X. Wang, T. Huang, J. Gonzalez, T. Darrell, and F. Yu, “Frustratingly simple few-shot object detection,” in Proceedings of the International Conference on Machine Learning . PMLR, 2020, pp. 9919–9928

work page 2020

[7] [7]

Defrcn: Decoupled faster r-cnn for few-shot object detection,

L. Qiao, Y . Zhao, Z. Li, X. Qiu, J. Wu, and C. Zhang, “Defrcn: Decoupled faster r-cnn for few-shot object detection,” in Proceedings of the IEEE International Conference on Computer Vision , 2021, pp. 8681–8690

work page 2021

[8] [8]

Few-shot object detection via association and discrimination,

Y . Cao, J. Wang, Y . Jin, T. Wu, K. Chen, Z. Liu, and D. Lin, “Few-shot object detection via association and discrimination,” Advances in Neural Information Processing Systems , vol. 34, pp. 16 570–16 581, 2021

work page 2021