首页|FSCIL-EACA:Few-Shot Class-Incremental Learning Network Based on Embedding Augmentation and Classifier Adaptation for Image Classification

FSCIL-EACA:Few-Shot Class-Incremental Learning Network Based on Embedding Augmentation and Classifier Adaptation for Image Classification

扫码查看
原文链接
  • 万方数据
The ability to learn incrementally is critical to the long-term operation of AI systems.Benefiting from the power of few-shot class-incremental learning(FSCIL),deep learning models can continuously recognize new class-es with only a few samples.The difficulty is that limited instances of new classes will lead to overfitting and exacer-bate the catastrophic forgetting of the old classes.Most previous works alleviate the above problems by imposing strong constraints on the model structure or parameters,but ignoring embedding network transferability and classifi-er adaptation(CA),failing to guarantee the efficient utilization of visual features and establishing relationships be-tween old and new classes.In this paper,we propose a simple and novel approach from two perspectives:embedding bias and classifier bias.The method learns an embedding augmented(EA)network with cross-class transfer and class-specific discriminative abilities based on self-supervised learning and modulated attention to alleviate embed-ding bias.Based on the adaptive incremental classifier learning scheme to realize incremental learning capability,guiding the adaptive update of prototypes and feature embeddings to alleviate classifier bias.We conduct extensive experiments on two popular natural image datasets and two medical datasets.The experiments show that our method is significantly better than the baseline and achieves state-of-the-art results.

Few-shot class-incremental learningEmbedding augmentationClassifier adaptationImage classi-fication

Ruru ZHANG、Haihong E、Meina SONG

展开 >

School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876,China

Education Department Information Network Engineering Research Center,Beijing University of Posts and Telecommunications,Beijing 100876,China

National Science Foundation of ChinaBeijing Natural Science FoundationEngineering Research Center of Information Networks,Ministry of Education

62176026M22009

2024

10.23919/cje.2022.00.396

电子学报(英文)

电子学报(英文)

CSTPCDEI
ISSN：1022-4653
年,卷(期)：2024.33(1)
  • 54