NSTL国家科技图书文献中心

1. Enhancing Recipe Retrieval with Foundation Models: A Data Augmentation Perspective NSTL国家科技图书文献中心

Fangzhou Song | Bin Zhu ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 111～127 - 共17页

摘要：Learning recipe and food image representation | in common embedding space is non-trivial but | crucial for cross-modal recipe retrieval. In this paper | , we propose a new perspective for this problem by | utilizing foundation models for data augmentation

关键词： Recipe retrieval | Data augmentation | Foundation models

2. SCOMatch: Alleviating Overtrusting in Open-Set Semi-supervised Learning NSTL国家科技图书文献中心

Zerun Wang | Liuyu Xiang ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 217～233 - 共17页

摘要：Open-set semi-supervised learning (OSSL | ) leverages practical open-set unlabeled data, comprising | both in-distribution (ID) samples from seen classes | and out-of-distribution (OOD) samples from unseen | classes, for semi-supervised learning (SSL). Prior OSSL

关键词： Open-set problem | Semi-supervised learning

3. On the Approximation Risk of Few-Shot Class-Incremental Learning NSTL国家科技图书文献中心

Xuan Wang | Zhong Ji ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 162～178 - 共17页

摘要：Few-Shot Class-Incremental Learning (FSCIL | ) aims to learn new concepts with few training samples | while preserving previously acquired knowledge | . Although promising performance has been achieved, there | remains an underexplored aspect regarding the basic

关键词： Few-Shot class-Incremental learning | Few-Shot learning | Class incremental learning | Approximation risk

4. StructLDM: Structured Latent Diffusion for 3D Human Generation NSTL国家科技图书文献中心

Tao Hu | Fangzhou Hong ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 363～381 - 共19页

摘要：Recent 3D human generative models have | achieved remarkable progress by learning 3D-aware GANs | from 2D images. However, existing 3D human generative | methods model humans in a compact 1D latent space | , ignoring the articulated structure and semantics of human

关键词： 3D human generation | Latent diffusion model

5. Diffusion Models as Optimizers for Efficient Planning in Offline RL NSTL国家科技图书文献中心

Renming Huang | Yunqiang Pei ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 1～17 - 共17页

摘要：Diffusion models have shown strong | competitiveness in offline reinforcement learning tasks by | formulating decision-making as sequential generation | . However, the practicality of these methods is limited | due to the lengthy inference processes they require

关键词： Reinforcement learning | Efficient planning | Diffusion model

6. INSTASTYLE: Inversion Noise of a Stylized Image is Secretly a Style Adviser NSTL国家科技图书文献中心

Xing Cui | Zekun Li ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 455～472 - 共18页

摘要：Stylized text-to-image generation focuses on | creating images from textual descriptions while adhering | to a style specified by reference images. However | , subtle style variations within different reference | images can hinder the model from accurately learning

关键词： Stylized image generation | Inversion noise | Signal-to-noise ratio | Prompt refinement

7. Long-CLIP: Unlocking the Long-Text Capability of CLIP NSTL国家科技图书文献中心

Beichen Zhang | Pan Zhang ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 310～325 - 共16页

摘要：Contrastive Language-Image Pre-training (CLIP | ) has been the cornerstone for zero-shot | classification, text-image retrieval, and text-image generation | by aligning image and text modalities. Despite its | widespread adoption, a significant limitation of CLIP lies

关键词： Multimodality | Zero-shot image classification | Text-Image retrieval | Text-to-Image generation

8. PapMOT: Exploring Adversarial Patch Attack Against Multiple Object Tracking NSTL国家科技图书文献中心

Jiahuan Long | Tingsong Jiang ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 128～144 - 共17页

摘要：Tracking multiple objects in a continuous | video stream is crucial for many computer vision tasks | . It involves detecting and associating objects with | their respective identities across successive frames | . Despite significant progress made in multiple object

关键词： Physical adversarial patches | Multiple object tracking | Evaluation metrics

9. How Many Are in This Image A Safety Evaluation Benchmark for Vision LLMs NSTL国家科技图书文献中心

Haoqin Tu | Chenhang Cui ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 37～55 - 共19页

摘要：This work focuses on benchmarking the | capabilities of vision large language models (VLLMs) in | visual reasoning. Different from prior studies, we | shift our focus from evaluating standard performance | to introducing a comprehensive safety evaluation

关键词： Vision language | Out-of-Distribution | Adversarial attack | Multimodal benchmark

10. Region-A ware Distribution Contrast: A Novel Approach to Multi-task Partially Supervised Learning NSTL国家科技图书文献中心

Meixuan Li | Tianyu Li ... - 《Computer Vision - ECCV 2024,Part LI》 - European Conference on Computer Vision - 2025, - 234～251 - 共18页

摘要：In this study, we address the intricate | challenge of multi-task dense prediction, encompassing | tasks such as semantic segmentation, depth estimation | , and surface normal estimation, particularly when | dealing with partially annotated data (MTPSL). The

摘要：Learning recipe and food image representation | in common embedding space is non-trivial but | crucial for cross-modal recipe retrieval. In this paper | , we propose a new perspective for this problem by | utilizing foundation models for data augmentation

关键词： Recipe retrieval | Data augmentation | Foundation models

摘要：Open-set semi-supervised learning (OSSL | ) leverages practical open-set unlabeled data, comprising | both in-distribution (ID) samples from seen classes | and out-of-distribution (OOD) samples from unseen | classes, for semi-supervised learning (SSL). Prior OSSL

关键词： Open-set problem | Semi-supervised learning

摘要：Few-Shot Class-Incremental Learning (FSCIL | ) aims to learn new concepts with few training samples | while preserving previously acquired knowledge | . Although promising performance has been achieved, there | remains an underexplored aspect regarding the basic

关键词： Few-Shot class-Incremental learning | Few-Shot learning | Class incremental learning | Approximation risk

摘要：Recent 3D human generative models have | achieved remarkable progress by learning 3D-aware GANs | from 2D images. However, existing 3D human generative | methods model humans in a compact 1D latent space | , ignoring the articulated structure and semantics of human

关键词： 3D human generation | Latent diffusion model

摘要：Diffusion models have shown strong | competitiveness in offline reinforcement learning tasks by | formulating decision-making as sequential generation | . However, the practicality of these methods is limited | due to the lengthy inference processes they require

关键词： Reinforcement learning | Efficient planning | Diffusion model

摘要：Stylized text-to-image generation focuses on | creating images from textual descriptions while adhering | to a style specified by reference images. However | , subtle style variations within different reference | images can hinder the model from accurately learning

关键词： Stylized image generation | Inversion noise | Signal-to-noise ratio | Prompt refinement

摘要：Contrastive Language-Image Pre-training (CLIP | ) has been the cornerstone for zero-shot | classification, text-image retrieval, and text-image generation | by aligning image and text modalities. Despite its | widespread adoption, a significant limitation of CLIP lies

关键词： Multimodality | Zero-shot image classification | Text-Image retrieval | Text-to-Image generation

摘要：Tracking multiple objects in a continuous | video stream is crucial for many computer vision tasks | . It involves detecting and associating objects with | their respective identities across successive frames | . Despite significant progress made in multiple object

关键词： Physical adversarial patches | Multiple object tracking | Evaluation metrics

摘要：This work focuses on benchmarking the | capabilities of vision large language models (VLLMs) in | visual reasoning. Different from prior studies, we | shift our focus from evaluating standard performance | to introducing a comprehensive safety evaluation

关键词： Vision language | Out-of-Distribution | Adversarial attack | Multimodal benchmark

摘要：In this study, we address the intricate | challenge of multi-task dense prediction, encompassing | tasks such as semantic segmentation, depth estimation | , and surface normal estimation, particularly when | dealing with partially annotated data (MTPSL). The

关键词： Multi-task learning | Partially supervised learning | Scene understanding | Contrastive learning

4008-161-200 800-990-8900

国家科技图书文献中心