作者: 戰(zhàn)勝 時(shí)間: 2025-3-21 23:16
,Constructing Balance from?Imbalance for?Long-Tailed Image Recognition,the separability of head-tail classes varies among different features with different inductive biases. Hence, our proposed model also provides a . method and paves the way for long-tailed . learning. Extensive experiments show that our method can boost the performance of state-of-the-arts of differe作者: Mosaic 時(shí)間: 2025-3-22 03:43 作者: 要塞 時(shí)間: 2025-3-22 07:28
,Worst Case Matters for?Few-Shot Recognition,o reduce the bias. Extensive experiments on benchmark datasets demonstrate the effectiveness of the proposed strategies, which outperforms current state-of-the-art methods with a significant margin in terms of not only average, but also worst-case accuracy.作者: gerontocracy 時(shí)間: 2025-3-22 10:40 作者: FLAT 時(shí)間: 2025-3-22 16:05 作者: FLAT 時(shí)間: 2025-3-22 18:16
,Dense Cross-Query-and-Support Attention Weighted Mask Aggregation for?Few-Shot Segmentation,dot-product attention in the Transformer architecture, DCAMA treats every query pixel as a token, computes its similarities with all support pixels, and predicts its segmentation label as an additive aggregation of all the support pixels’ labels—weighted by the similarities. Based on the unique form作者: Fecal-Impaction 時(shí)間: 2025-3-23 00:35
,Rethinking Clustering-Based Pseudo-Labeling for?Unsupervised Meta-Learning, alleviate the limited diversity problem. Finally, our approach is also model-agnostic and can easily be integrated into existing supervised methods. To demonstrate its generalization ability, we integrate it into two representative algorithms: MAML and EP. The results on three main few-shot benchma作者: 褪色 時(shí)間: 2025-3-23 03:17 作者: Ardent 時(shí)間: 2025-3-23 07:36 作者: INCUR 時(shí)間: 2025-3-23 13:38
,Meta-Learning with?Less Forgetting on?Large-Scale Non-Stationary Task Distributions,tributions sequentially arrive with some ORDER), to tackle these two major challenges. Specifically, our ORDER introduces a novel mutual information regularization to robustify the model with unlabeled OOD data and adopts an optimal transport regularization to remember previously learned knowledge i作者: crockery 時(shí)間: 2025-3-23 15:04
,DnA: Improving Few-Shot Transfer Learning with?Low-Rank Decomposition and?Alignment,the low-rank subspace), and the extra flexibility to absorb the new out-of-the-domain knowledge (via freeing the sparse residual). Our resultant framework, termed Decomposition-and-Alignment (.), significantly improves the few-shot transfer performance of the SS pre-trained model to downstream tasks作者: 打包 時(shí)間: 2025-3-23 18:22 作者: 無效 時(shí)間: 2025-3-23 22:10
,Open-World Semantic Segmentation via?Contrasting and?Clustering Vision-Language Embedding,mage encoder is jointly trained with a vision-based contrasting and a cross-modal contrasting, which encourage the visual embeddings to preserve both fine-grained semantics and high-level category information that are crucial for the segmentation task. Furthermore, an online clustering head is devis作者: Essential 時(shí)間: 2025-3-24 05:49 作者: 火海 時(shí)間: 2025-3-24 07:34 作者: OTTER 時(shí)間: 2025-3-24 12:16 作者: aristocracy 時(shí)間: 2025-3-24 18:04 作者: Servile 時(shí)間: 2025-3-24 19:07 作者: 易碎 時(shí)間: 2025-3-25 02:13 作者: 細(xì)節(jié) 時(shí)間: 2025-3-25 06:26 作者: Synthesize 時(shí)間: 2025-3-25 11:30 作者: Stable-Angina 時(shí)間: 2025-3-25 14:42
Comparative Ecocriticism: An Introduction,dot-product attention in the Transformer architecture, DCAMA treats every query pixel as a token, computes its similarities with all support pixels, and predicts its segmentation label as an additive aggregation of all the support pixels’ labels—weighted by the similarities. Based on the unique form作者: 拘留 時(shí)間: 2025-3-25 16:34
https://doi.org/10.1007/978-3-319-46425-1 alleviate the limited diversity problem. Finally, our approach is also model-agnostic and can easily be integrated into existing supervised methods. To demonstrate its generalization ability, we integrate it into two representative algorithms: MAML and EP. The results on three main few-shot benchma作者: Exuberance 時(shí)間: 2025-3-25 21:08 作者: 含糊 時(shí)間: 2025-3-26 02:59 作者: 粗糙 時(shí)間: 2025-3-26 06:54
Pedro Duarte,Jan Marcin Weslawski,Haakon Hoptributions sequentially arrive with some ORDER), to tackle these two major challenges. Specifically, our ORDER introduces a novel mutual information regularization to robustify the model with unlabeled OOD data and adopts an optimal transport regularization to remember previously learned knowledge i作者: Salivary-Gland 時(shí)間: 2025-3-26 10:23 作者: 外表讀作 時(shí)間: 2025-3-26 13:29
Olga Pavlova,Sebastian Gerland,Haakon Hopr model improves performance on few-shot classification and detection tasks, achieving a tangible improvement over several baseline models. This includes state-of-the-art results on four few-shot classification benchmarks: .-ImageNet, .-ImageNet, CUB and FC100 and competitive results on a few-shot d作者: 合法 時(shí)間: 2025-3-26 17:39 作者: 仔細(xì)檢查 時(shí)間: 2025-3-26 23:08 作者: 針葉類的樹 時(shí)間: 2025-3-27 04:47
Abhishek Kathuria,Prasanna P. Karhade generate individual location-specific supervision for guiding each patch token. This location-specific supervision tells the ViT which patch tokens are similar or dissimilar and thus accelerates token dependency learning. Moreover, it models the local semantics in each patch token to improve the ob作者: 分解 時(shí)間: 2025-3-27 07:01
Conference proceedings 2022ning; object recognition; image classification; image processing; object detection; semantic segmentation; human pose estimation; 3d reconstruction; stereo vision; computational photography; neural networks; image coding; image reconstruction; object recognition; motion estimation..作者: 過份 時(shí)間: 2025-3-27 12:11
0302-9743 ruction; stereo vision; computational photography; neural networks; image coding; image reconstruction; object recognition; motion estimation..978-3-031-20043-4978-3-031-20044-1Series ISSN 0302-9743 Series E-ISSN 1611-3349 作者: Nonporous 時(shí)間: 2025-3-27 16:13 作者: Cryptic 時(shí)間: 2025-3-27 21:04 作者: Banister 時(shí)間: 2025-3-27 22:58
Jaehwan Lee,Byungjoon Yoo,Moonkyoung Jange-pooled support embedding. We also propose a Transformer Relation Head (TRH), equipped with higher-order representations, which encodes correlations between query regions and the entire support set, while being sensitive to the positional variability of object instances. Our model achieves state-of-the-art results on PASCAL VOC, FSOD, and COCO.作者: 說不出 時(shí)間: 2025-3-28 03:56 作者: Armada 時(shí)間: 2025-3-28 08:54 作者: slipped-disk 時(shí)間: 2025-3-28 13:22 作者: 脾氣暴躁的人 時(shí)間: 2025-3-28 15:25 作者: 憎惡 時(shí)間: 2025-3-28 22:44 作者: Fortuitous 時(shí)間: 2025-3-29 00:17
,tSF: Transformer-Based Semantic Filter for?Few-Shot Learning,een (novel) labeled samples. Most feature embedding modules in recent FSL methods are specially designed for corresponding learning tasks (e.g., classification, segmentation, and object detection), which limits the . of embedding features. To this end, we propose a light and universal module named t作者: ironic 時(shí)間: 2025-3-29 05:53
,Adversarial Feature Augmentation for?Cross-domain Few-Shot Classification,sting methods based on meta-learning predict novel-class labels for (target domain) testing tasks via meta knowledge learned from (source domain) training tasks of base classes. However, most existing works may fail to generalize to novel classes due to the probably large domain discrepancy across d作者: headlong 時(shí)間: 2025-3-29 10:42
,Constructing Balance from?Imbalance for?Long-Tailed Image Recognition,ail) classes severely skews the data-driven deep neural networks. Previous methods tackle with data imbalance from the viewpoints of data distribution, feature space, and model design, etc. In this work, instead of directly learning a recognition model, we suggest confronting the bottleneck of head-作者: 猛烈責(zé)罵 時(shí)間: 2025-3-29 12:08
,On Multi-Domain Long-Tailed Recognition, Imbalanced Domain Generalization and?Beyond,om the same data distribution. However, natural data can originate from distinct domains, where a minority class in one domain could have abundant instances from other domains. We formalize the task of Multi-Domain Long-Tailed Recognition (MDLT), which learns from multi-domain imbalanced data, addre作者: 頑固 時(shí)間: 2025-3-29 17:02 作者: 單調(diào)性 時(shí)間: 2025-3-29 20:43 作者: annexation 時(shí)間: 2025-3-30 02:28
,Exploring Hierarchical Graph Representation for?Large-Scale Zero-Shot Image Classification,ands of categories as in the ImageNet-21K benchmark. At this scale, especially with many fine-grained categories included in ImageNet-21K, it is critical to learn quality visual semantic representations that are discriminative enough to recognize unseen classes and distinguish them from seen ones. W作者: 翻動(dòng) 時(shí)間: 2025-3-30 06:22
Doubly Deformable Aggregation of Covariance Matrices for Few-Shot Segmentation, task, the main challenge is how to accurately measure the semantic correspondence between the support and query samples with limited training data. To address this problem, we propose to aggregate the learnable covariance matrices with a deformable 4D Transformer to effectively predict the segmenta作者: Irremediable 時(shí)間: 2025-3-30 11:17
,Dense Cross-Query-and-Support Attention Weighted Mask Aggregation for?Few-Shot Segmentation,a few annotated support images of the target class. A key to this challenging task is to fully utilize the information in the support images by exploiting fine-grained correlations between the query and support images. However, most existing approaches either compressed the support information into 作者: 連接 時(shí)間: 2025-3-30 14:05 作者: Keshan-disease 時(shí)間: 2025-3-30 18:49
,CLASTER: Clustering with?Reinforcement Learning for?Zero-Shot Action Recognition,n seen classes which generalizes well to instances of unseen classes, without losing discriminability between classes. Neural networks are able to model highly complex boundaries between visual classes, which explains their success as supervised models. However, in Zero-Shot learning, these highly s作者: bourgeois 時(shí)間: 2025-3-30 21:50 作者: 拍翅 時(shí)間: 2025-3-31 04:26 作者: MIRE 時(shí)間: 2025-3-31 07:45
,DnA: Improving Few-Shot Transfer Learning with?Low-Rank Decomposition and?Alignment,ver, when transferring such representations to downstream tasks with domain shifts, the performance degrades compared to its supervised counterpart, especially at the few-shot regime. In this paper, we proposed to boost the transferability of the self-supervised pre-trained models on cross-domain ta作者: 宇宙你 時(shí)間: 2025-3-31 10:27
,Learning Instance and?Task-Aware Dynamic Kernels for?Few-Shot Learning,le way of achieving few-shot learning is to realize a model that can rapidly adapt to the context of a given task. Dynamic networks have been shown capable of learning content-adaptive parameters efficiently, making them suitable for few-shot learning. In this paper, we propose to learn the dynamic 作者: forecast 時(shí)間: 2025-3-31 17:05
,Open-World Semantic Segmentation via?Contrasting and?Clustering Vision-Language Embedding,cent zero-shot segmentation attracts a lot of attention by exploring the relationships between unseen and seen object categories, yet requiring large amounts of densely-annotated data with diverse base classes. In this paper, we propose a new open-world semantic segmentation pipeline that makes the 作者: RAFF 時(shí)間: 2025-3-31 18:30 作者: Accrue 時(shí)間: 2025-3-31 21:52
,Time-rEversed DiffusioN tEnsor Transformer: A New TENET of?Few-Shot Object Detection,ult in information loss; and/or (ii) discard position information that can help detect object instances. Consequently, such pipelines are sensitive to large intra-class appearance and geometric variations between support and query images. To address these drawbacks, we propose a Time-rEversed diffus作者: 婚姻生活 時(shí)間: 2025-4-1 02:29
,Self-Promoted Supervision for?Few-Shot Transformer, the same few-shot learning frameworks, replacing the widely used CNN feature extractor with a ViT model often severely impairs few-shot classification performance. Moreover, our empirical study shows that in the absence of inductive bias, ViTs often learn the low-qualified token dependencies under 作者: GORGE 時(shí)間: 2025-4-1 09:05
Computer Vision – ECCV 2022978-3-031-20044-1Series ISSN 0302-9743 Series E-ISSN 1611-3349 作者: 確定方向 時(shí)間: 2025-4-1 10:50
