NSTL国家科技图书文献中心

1. Versatile Incremental Learning: Towards Class and Domain-Agnostic Incremental Learning NSTL国家科技图书文献中心

Min-Yeong Park | Jae-Ho Lee ... - 《Computer Vision - ECCV 2024,Part XXXI》 - European Conference on Computer Vision - 2025, - 271～288 - 共18页

摘要：Incremental Learning (IL) aims to accumulate | scenario, named Versatile Incremental* Learning (VIL), in* | , named Incremental* Classifier with Adaptation Shift* | effectively. Moreover, we introduce an Incremental | knowledge from sequential input tasks while overcoming

关键词： Incremental learning | Real-world scenario | Adaptation control | Incremental classifier

2. Future-proofing class-incremental learning NSTL国家科技图书文献中心

Quentin Jodelet | Xin Liu ... - 《Machine Vision and Applications》 - 2025,36(1) - 16.1～16.16 - 共16页

摘要：Exemplar-free class incremental* learning is a* | available during the first incremental* step. To overcome* | -the-art methods for exemplar-free class incremental | learning, especially in the most difficult settings where | the first incremental* step only contains few classes*

关键词： Class incremental learning | Continual learning | Image classification | Image generation

3. Adaptive Decoupled Prompting for Class Incremental Learning NSTL国家科技图书文献中心

Fanhao Zhang | Shiye Wang ... - 《Pattern Recognition and Computer Vision,Part IX》 - Chinese Conference on Pattern Recognition and Computer Vision - 2025, - 554～568 - 共15页

摘要： for Class Incremental* Learning) in comparison to the* | related class-incremental* learning methods.* | Continual learning* has garnered significant* | adaptive decoupled prompting method for class incremental | learning. Specifically, we design an adaptive prompt

关键词： Prompt learning | Incremental learning | Catastrophic forgetting

4. PILoRA: Prototype Guided Incremental LoRA for Federated Class-Incremental Learning NSTL国家科技图书文献中心

Haiyang Guo | Fei Zhu ... - 《Computer Vision - ECCV 2024,Part LXV》 - European Conference on Computer Vision - 2025, - 141～159 - 共19页

摘要： view incremental learning as the process of learning | Existing federated learning* methods have* | effectively dealt with decentralized learning* in scenarios* | PILoRA. On the one hand, we adopt prototype learning* to* | LoRA parameters. Accordingly, we propose Incremental

关键词： Federated learning | Class incremental learning

5. Federated Incremental Learning algorithm based on Topological Data Analysis NSTL国家科技图书文献中心

Hu, Kai | Gong, Sheng ... - 《Pattern Recognition》 - 2025,158 - 共14页

摘要： incremental learning is an adaptive machine learning method | Incremental Learning (FIL) algorithms brings together their | incremental learning still faces two main challenges: (1 | % respectively for the federated incremental* learning model* | Federated learning is a distributed learning

关键词： Federated learning | Incremental learning | Topological Data Analysis

6. DiffClass: Diffusion-Based Class Incremental Learning NSTL国家科技图书文献中心

Zichong Meng | Jie Zhang ... - 《Computer Vision - ECCV 2024,Part LXXXVII》 - European Conference on Computer Vision - 2025, - 142～159 - 共18页

摘要：Class Incremental* Learning (CIL) is* | enhance model stability during incremental* training* | challenging due to catastrophic forgetting. On top of that | , exemplar-free CIL is even more challenging due to | forbidden access to data of previous tasks. Recent

关键词： Class incremental learning | Exemplar free | Diffusion model

7. iNeMo: Incremental Neural Mesh Models for Robust Class-Incremental Learning NSTL国家科技图书文献中心

Tom Fischer | Yaoyao Liu ... - 《Computer Vision - ECCV 2024,Part LXXVII》 - European Conference on Computer Vision - 2025, - 357～374 - 共18页

摘要： the first incremental* learning approach for pose* | learning models only initially and on fixed datasets. A | property in a continual learning* setting, we propose* | incremental neural mesh models that can be extended with new | Different from human nature, it is still

关键词： Class-incremental learning | 3D pose estimation

8. SRIL: Selective Regularization for Class-Incremental Learning NSTL国家科技图书文献中心

Jisu Han | Jaemin Na ... - 《Computer Vision - ACCV 2024,Part VIII》 - Asian Conference on Computer Vision - 2025, - 351～367 - 共17页

摘要：-Incremental* Learning aims to create an integrated model that* | lifespan. However, deep learning* models suffer from a* | and enable exploratory learning. We validate the | Human intelligence gradually accepts new | information and accumulates knowledge throughout the

关键词： Class incremental learning | Knowledge distillation | Weight interpolation

9. Adaptive Knowledge Matching for Exemplar-Free Class-Incremental Learning NSTL国家科技图书文献中心

Runhang Chen | Xiao-Yuan Jing ... - 《Pattern Recognition and Computer Vision,Part III》 - Chinese Conference on Pattern Recognition and Computer Vision - 2025, - 289～303 - 共15页

摘要：Exemplar-free class-incremental learning | (EFCIL) presents a significant challenge, requiring | models to learn tasks sequentially without accessing | data from previous tasks. This challenge is | exacerbated when the initial dataset is insufficient for

关键词： Class-Incremental learning | Exemplar-Free class-Incremental learning | Knowledge distillation

10. Non-exemplar Domain Incremental Learning via Cross-Domain Concept Integration NSTL国家科技图书文献中心

Qiang Wang | Yuhang He ... - 《Computer Vision - ECCV 2024,Part XLIV》 - European Conference on Computer Vision - 2025, - 144～162 - 共19页

摘要：. Therefore, Non-Exemplar Domain Incremental* Learning (NEDIL* | Existing approaches to Domain Incremental | Learning (DIL) address catastrophic forgetting by storing | (DSA) module is proposed for each incremental* domain* | and rehearsing exemplars from old domains. However

摘要：Incremental Learning (IL) aims to accumulate | scenario, named Versatile Incremental* Learning (VIL), in* | , named Incremental* Classifier with Adaptation Shift* | effectively. Moreover, we introduce an Incremental | knowledge from sequential input tasks while overcoming

关键词： Incremental learning | Real-world scenario | Adaptation control | Incremental classifier

摘要：Exemplar-free class incremental* learning is a* | available during the first incremental* step. To overcome* | -the-art methods for exemplar-free class incremental | learning, especially in the most difficult settings where | the first incremental* step only contains few classes*

关键词： Class incremental learning | Continual learning | Image classification | Image generation

摘要： for Class Incremental* Learning) in comparison to the* | related class-incremental* learning methods.* | Continual learning* has garnered significant* | adaptive decoupled prompting method for class incremental | learning. Specifically, we design an adaptive prompt

关键词： Prompt learning | Incremental learning | Catastrophic forgetting

摘要： view incremental learning as the process of learning | Existing federated learning* methods have* | effectively dealt with decentralized learning* in scenarios* | PILoRA. On the one hand, we adopt prototype learning* to* | LoRA parameters. Accordingly, we propose Incremental

关键词： Federated learning | Class incremental learning

关键词： Federated learning | Incremental learning | Topological Data Analysis

摘要：Class Incremental* Learning (CIL) is* | enhance model stability during incremental* training* | challenging due to catastrophic forgetting. On top of that | , exemplar-free CIL is even more challenging due to | forbidden access to data of previous tasks. Recent

关键词： Class incremental learning | Exemplar free | Diffusion model

摘要： the first incremental* learning approach for pose* | learning models only initially and on fixed datasets. A | property in a continual learning* setting, we propose* | incremental neural mesh models that can be extended with new | Different from human nature, it is still

关键词： Class-incremental learning | 3D pose estimation

摘要：-Incremental* Learning aims to create an integrated model that* | lifespan. However, deep learning* models suffer from a* | and enable exploratory learning. We validate the | Human intelligence gradually accepts new | information and accumulates knowledge throughout the

关键词： Class incremental learning | Knowledge distillation | Weight interpolation

摘要：Exemplar-free class-incremental learning | (EFCIL) presents a significant challenge, requiring | models to learn tasks sequentially without accessing | data from previous tasks. This challenge is | exacerbated when the initial dataset is insufficient for

关键词： Class-Incremental learning | Exemplar-Free class-Incremental learning | Knowledge distillation

摘要：. Therefore, Non-Exemplar Domain Incremental* Learning (NEDIL* | Existing approaches to Domain Incremental | Learning (DIL) address catastrophic forgetting by storing | (DSA) module is proposed for each incremental* domain* | and rehearsing exemplars from old domains. However

关键词： Domain incremental learning | Non-exemplar | Vision transformer

4008-161-200 800-990-8900

国家科技图书文献中心

摘要：Incremental Learning (IL) aims to accumulate | scenario, named Versatile Incremental Learning (VIL), in | , named Incremental Classifier with Adaptation Shift | effectively. Moreover, we introduce an Incremental | knowledge from sequential input tasks while overcoming

关键词： Incremental learning | Real-world scenario | Adaptation control | Incremental classifier

摘要：Exemplar-free class incremental learning is a | available during the first incremental step. To overcome | -the-art methods for exemplar-free class incremental | learning, especially in the most difficult settings where | the first incremental step only contains few classes

关键词： Class incremental learning | Continual learning | Image classification | Image generation

摘要： for Class Incremental Learning) in comparison to the | related class-incremental learning methods. | Continual learning has garnered significant | adaptive decoupled prompting method for class incremental | learning. Specifically, we design an adaptive prompt

关键词： Prompt learning | Incremental learning | Catastrophic forgetting

摘要： view incremental learning as the process of learning | Existing federated learning methods have | effectively dealt with decentralized learning in scenarios | PILoRA. On the one hand, we adopt prototype learning to | LoRA parameters. Accordingly, we propose Incremental

关键词： Federated learning | Class incremental learning

摘要： incremental learning is an adaptive machine learning method | Incremental Learning (FIL) algorithms brings together their | incremental learning still faces two main challenges: (1 | % respectively for the federated incremental learning model | Federated learning is a distributed learning

关键词： Federated learning | Incremental learning | Topological Data Analysis

摘要：Class Incremental Learning (CIL) is | enhance model stability during incremental training | challenging due to catastrophic forgetting. On top of that | , exemplar-free CIL is even more challenging due to | forbidden access to data of previous tasks. Recent

关键词： Class incremental learning | Exemplar free | Diffusion model

摘要： the first incremental learning approach for pose | learning models only initially and on fixed datasets. A | property in a continual learning setting, we propose | incremental neural mesh models that can be extended with new | Different from human nature, it is still

关键词： Class-incremental learning | 3D pose estimation

摘要：-Incremental Learning aims to create an integrated model that | lifespan. However, deep learning models suffer from a | and enable exploratory learning. We validate the | Human intelligence gradually accepts new | information and accumulates knowledge throughout the

关键词： Class incremental learning | Knowledge distillation | Weight interpolation

摘要：Exemplar-free class-incremental learning | (EFCIL) presents a significant challenge, requiring | models to learn tasks sequentially without accessing | data from previous tasks. This challenge is | exacerbated when the initial dataset is insufficient for

关键词： Class-Incremental learning | Exemplar-Free class-Incremental learning | Knowledge distillation

摘要：. Therefore, Non-Exemplar Domain Incremental Learning (NEDIL | Existing approaches to Domain Incremental | Learning (DIL) address catastrophic forgetting by storing | (DSA) module is proposed for each incremental domain | and rehearsing exemplars from old domains. However

关键词： Domain incremental learning | Non-exemplar | Vision transformer

4008-161-200 800-990-8900

国家科技图书文献中心

摘要：Incremental Learning (IL) aims to accumulate | scenario, named Versatile Incremental* Learning (VIL), in* | , named Incremental* Classifier with Adaptation Shift* | effectively. Moreover, we introduce an Incremental | knowledge from sequential input tasks while overcoming

摘要：Exemplar-free class incremental* learning is a* | available during the first incremental* step. To overcome* | -the-art methods for exemplar-free class incremental | learning, especially in the most difficult settings where | the first incremental* step only contains few classes*

摘要： for Class Incremental* Learning) in comparison to the* | related class-incremental* learning methods.* | Continual learning* has garnered significant* | adaptive decoupled prompting method for class incremental | learning. Specifically, we design an adaptive prompt

摘要： view incremental learning as the process of learning | Existing federated learning* methods have* | effectively dealt with decentralized learning* in scenarios* | PILoRA. On the one hand, we adopt prototype learning* to* | LoRA parameters. Accordingly, we propose Incremental

摘要：Class Incremental* Learning (CIL) is* | enhance model stability during incremental* training* | challenging due to catastrophic forgetting. On top of that | , exemplar-free CIL is even more challenging due to | forbidden access to data of previous tasks. Recent

摘要： the first incremental* learning approach for pose* | learning models only initially and on fixed datasets. A | property in a continual learning* setting, we propose* | incremental neural mesh models that can be extended with new | Different from human nature, it is still

摘要：-Incremental* Learning aims to create an integrated model that* | lifespan. However, deep learning* models suffer from a* | and enable exploratory learning. We validate the | Human intelligence gradually accepts new | information and accumulates knowledge throughout the

摘要：. Therefore, Non-Exemplar Domain Incremental* Learning (NEDIL* | Existing approaches to Domain Incremental | Learning (DIL) address catastrophic forgetting by storing | (DSA) module is proposed for each incremental* domain* | and rehearsing exemplars from old domains. However