TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Class Incremental Learning	cifar100	iCaRL	10-stage average accuracy	63.24	# 3
Incremental Learning	CIFAR-100 - 50 classes + 10 steps of 5 classes	iCaRL*	Average Incremental Accuracy	52.57	# 13
Incremental Learning	CIFAR-100 - 50 classes + 2 steps of 25 classes	iCaRL	Average Incremental Accuracy	71.33	# 4
Incremental Learning	CIFAR-100 - 50 classes + 5 steps of 10 classes	iCaRL*	Average Incremental Accuracy	57.17	# 12
Incremental Learning	CIFAR-100-B0(5steps of 20 classes)	iCaRL	Average Incremental Accuracy	71.14	# 6
Incremental Learning	ImageNet100 - 10 steps	iCaRL	Average Incremental Accuracy Top-5	83.60	# 10
Incremental Learning	ImageNet100 - 10 steps	iCaRL	Final Accuracy Top-5	63.80	# 10
Incremental Learning	ImageNet100 - 10 steps	iCaRL	# M Params	11.22	# 2
Incremental Learning	ImageNet-100 - 50 classes + 5 steps of 10 classes	iCaRL*	Average Incremental Accuracy	65.56	# 5
Incremental Learning	ImageNet - 10 steps	iCaRL	Average Incremental Accuracy	38.40	# 8
Incremental Learning	ImageNet - 10 steps	iCaRL	Final Accuracy	22.70	# 5
Incremental Learning	ImageNet - 10 steps	iCaRL	Average Incremental Accuracy Top-5	63.70	# 7
Incremental Learning	ImageNet - 10 steps	iCaRL	Final Accuracy Top-5	44.00	# 7
Incremental Learning	ImageNet - 10 steps	iCaRL	# M Params	11.68	# 2

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-imagenet100-10-steps)](https://paperswithcode.com/sota/incremental-learning-on-imagenet100-10-steps?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/class-incremental-learning-on-cifar100)](https://paperswithcode.com/sota/class-incremental-learning-on-cifar100?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-cifar-100-50-classes-4)](https://paperswithcode.com/sota/incremental-learning-on-cifar-100-50-classes-4?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-imagenet-100-50-3)](https://paperswithcode.com/sota/incremental-learning-on-imagenet-100-50-3?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-cifar-100-b0-5steps)](https://paperswithcode.com/sota/incremental-learning-on-cifar-100-b0-5steps?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-imagenet-10-steps)](https://paperswithcode.com/sota/incremental-learning-on-imagenet-10-steps?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-cifar-100-50-classes-3)](https://paperswithcode.com/sota/incremental-learning-on-cifar-100-50-classes-3?p=icarl-incremental-classifier-and)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/icarl-incremental-classifier-and/incremental-learning-on-cifar-100-50-classes-2)](https://paperswithcode.com/sota/incremental-learning-on-cifar-100-50-classes-2?p=icarl-incremental-classifier-and)`

iCaRL: Incremental Classifier and Representation Learning

CVPR 2017 · Sylvestre-Alvise Rebuffi, Alexander Kolesnikov, Georg Sperl, Christoph H. Lampert ·

A major open problem on the road to artificial intelligence is the development of incrementally learning systems that learn about more and more concepts over time from a stream of data. In this work, we introduce a new training strategy, iCaRL, that allows learning in such a class-incremental way: only the training data for a small number of classes has to be present at the same time and new classes can be added progressively. iCaRL learns strong classifiers and a data representation simultaneously. This distinguishes it from earlier works that were fundamentally limited to fixed data representations and therefore incompatible with deep learning architectures. We show by experiments on CIFAR-100 and ImageNet ILSVRC 2012 data that iCaRL can learn many classes incrementally over a long period of time where other strategies quickly fail.

PDF Abstract CVPR 2017 PDF CVPR 2017 Abstract

Code

Add Remove Mark official

srebuffi/iCaRL official

238

ContinualAI/avalanche

1,664

g-u-n/pycil

687

mmasana/FACIL

496

yaoyao-liu/mnemonics

454

See all 9 implementations

Tasks

Add Remove

Class Incremental Learning

Incremental Learning

Representation Learning

Datasets

CIFAR-100

Results from the Paper

Edit

Ranked #2 on Incremental Learning on ImageNet100 - 10 steps (# M Params metric)

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Class Incremental Learning	cifar100	iCaRL	10-stage average accuracy	63.24	# 3	Compare
Incremental Learning	CIFAR-100 - 50 classes + 2 steps of 25 classes	iCaRL	Average Incremental Accuracy	71.33	# 4	Compare
Incremental Learning	CIFAR-100-B0(5steps of 20 classes)	iCaRL	Average Incremental Accuracy	71.14	# 6	Compare
Incremental Learning	ImageNet-100 - 50 classes + 5 steps of 10 classes	iCaRL*	Average Incremental Accuracy	65.56	# 5	Compare
Incremental Learning	ImageNet - 10 steps	iCaRL	Average Incremental Accuracy	38.40	# 8	Compare
			Final Accuracy	22.70	# 5	Compare
			Average Incremental Accuracy Top-5	63.70	# 7	Compare
			Final Accuracy Top-5	44.00	# 7	Compare
			# M Params	11.68	# 2	Compare

Results from Other Papers

Task	Dataset	Model	Metric Name	Metric Value	Rank	Compare
Incremental Learning	CIFAR-100 - 50 classes + 10 steps of 5 classes	iCaRL*	Average Incremental Accuracy	52.57	# 13	See all
Incremental Learning	CIFAR-100 - 50 classes + 5 steps of 10 classes	iCaRL*	Average Incremental Accuracy	57.17	# 12	See all
Incremental Learning	ImageNet100 - 10 steps	iCaRL	Average Incremental Accuracy Top-5	83.60	# 10	See all
			Final Accuracy Top-5	63.80	# 10	See all
			# M Params	11.22	# 2	See all

Methods

Add Remove

No methods listed for this paper. Add relevant methods here

Edit Social Preview

iCaRL: Incremental Classifier and Representation Learning

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit