Few-Shot Class-Incremental Learning

Tao, Xiaoyu; Hong, Xiaopeng; Chang, Xinyuan; Dong, Songlin; Wei, Xing; Gong, Yi

doi:10.1109/cvpr42600.2020.01220

Cited by 312 publications

(287 citation statements)

References 15 publications

Supporting

Mentioning

286

Contrasting

Order By: Relevance

“…The fewshot learning setting results of supervised iCaRL and NCM are also reported. We also compare our results with Ft-CNN [1] and Joint-CNN [1]. In Joint-CNN, all labeled examples from previous sessions take part in the training process in the current session.…”

Section: Resultsmentioning

confidence: 99%

“…We conducted extensive experiments on three popular image classification databases, i.e., CIFAR100 [22], miniImageNet [10], and CUB200 [23]. For a direct comparison, we followed the same split protocol as in [1] which is the baseline of FSCIL. We implemented SSFSCIL by self-training based on two incremental learning methods, i.e., iCaRL [17] and NCM [16].…”

Section: Methodsmentioning

confidence: 99%

“…Therefore, our brain constantly receives and learns new concepts based on a few samples and updates the boundaries between the learned ones. This competence is termed as few-shot class-incremental learning (FSCIL) [1]. It is also crucial for deep models to acquire this ability in some real applications, such as medical image analysis [2,3,4] and autonomous driving vehicles [5,6,7].…”

Section: Introductionmentioning

confidence: 99%

“…It aims to continuously learn new categories with limited labeled data arriving in the current session while avoid catastrophic forgetting the old categories of previous sessions. FSCIL has been rarely explored [1]. In this paper, we propose a novel Semi-Supervised Few-Shot Class-Incremental Learning (SSFSCIL) to incrementally learn novel categories with limited labeled and a larger number of unlabeled samples.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Semi-Supervised Few-Shot Class-Incremental Learning

Cui¹,

Xiong²,

Tavakolian

et al. 2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

The capability of incrementally learning new classes and learning from a few examples is one of the hallmarks of human intelligence. It is crucial to endow a practical recognition system with such ability. Therefore, in this paper, we conduct pioneering work and focus on a challenging yet practical Semi-Supervised Few-Shot Class-Incremental Learning (SSFSCIL) problem, which requires CNN models incrementally learn new classes from very few labeled samples and a large number of unlabeled samples, without forgetting the previously learned ones. To address this problem, a simple and efficient solution for SSFSCIL is proposed to learn novel categories using a self-training strategy in a semi-supervised manner and avoid catastrophic forgetting by distillation-based methods. Our extensive experiments on CIFAR100, miniImageNet and CUB200 datasets demonstrate the promising performance of our proposed method, and define baselines in this new research direction.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Semi-Supervised Few-Shot Class-Incremental Learning

Cui¹,

Xiong²,

Tavakolian

et al. 2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

show abstract

“…Recently, the novel research field of few-shot continual learning (few-shot incremental learning, low-shot learning) combines the strengths of the aforementioned approaches and aims to continuously expand the capability of a classifier based on only few data at inference time [25][26][27][28]. This enables fast and interactive model updates by end-users.…”

Section: Introductionmentioning

confidence: 99%

Few-Shot Continual Learning for Audio Classification

Wang

Bryan

Cartwright

et al. 2021

ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Supervised learning for audio classification typically imposes a fixed class vocabulary, which can be limiting for real-world applications where the target class vocabulary is not known a priori or changes dynamically. In this work, we introduce a few-shot continual learning framework for audio classification, where we can continuously expand a trained base classifier to recognize novel classes based on only few labeled data at inference time. This enables fast and interactive model updates by end-users with minimal human effort. To do so, we leverage the dynamic few-shot learning technique and adapt it to a challenging multi-label audio classification scenario. We incorporate a recent state-of-the-art audio feature extraction model as a backbone and perform a comparative analysis of our approach on two popular audio datasets (ESC-50 and AudioSet). We conduct an in-depth evaluation to illustrate the complexities of the problem and show that, while there is still room for improvement, our method outperforms three baselines on novel class detection while maintaining its performance on base classes.

show abstract

Remembering for the right reasons: Explanations reduce catastrophic forgetting

et al. 2021

View full text Add to dashboard Cite

The goal of continual learning (CL) is to learn a sequence of tasks without suffering from the phenomenon of catastrophic forgetting. Previous work has shown that leveraging memory in the form of a replay buffer can reduce performance degradation on prior tasks. We hypothesize that forgetting can be further reduced when the model is encouraged to remember the evidence for previously made decisions. As a first step towards exploring this hypothesis, we propose a simple novel training paradigm, called Remembering for the Right Reasons (RRR), that additionally stores visual model explanations for each example in the buffer and ensures the model has "the right reasons" for its predictions by encouraging its explanations to remain consistent with those used to make decisions at training time. Without this constraint, there is a drift in explanations and increase in forgetting as conventional continual learning algorithms learn new tasks. We demonstrate how RRR can be easily added to any memory or regularization-based approach and results in reduced forgetting, and more importantly, improved model explanations. We have evaluated our approach in the standard and fewshot settings and observed a consistent improvement across various CL approaches using different architectures and techniques to generate model explanations and demonstrated our approach showing a promising connection between explainability and continual learning. Our code is available at https://github.com/ SaynaEbrahimi/Remembering-for-the-Right-Reasons.

show abstract

Few-Shot Class-Incremental Learning

Cited by 312 publications

References 15 publications

Semi-Supervised Few-Shot Class-Incremental Learning

Semi-Supervised Few-Shot Class-Incremental Learning

Few-Shot Continual Learning for Audio Classification

Remembering for the right reasons: Explanations reduce catastrophic forgetting

Contact Info

Product

Resources

About