Entropy-based active learning for object recognition

Holub, Alex D.; Perona, Pietro; Burl, Michael C.

doi:10.1109/cvprw.2008.4563068

Cited by 215 publications

(137 citation statements)

References 10 publications

(11 reference statements)

Supporting

Mentioning

136

Contrasting

Unclassified

Order By: Relevance

“…In our work, we use the one-versus-one SVM formulation, and allow the addition of a variable number of examples at each iteration. Holub et al [9] recently proposed a multi-class active learning method. Their methods selects examples from the active pool, whose addition to the training set minimizes the expected entropy of the system.…”

Section: Previous Workmentioning

confidence: 99%

Multi-class active learning for image classification

Joshi

Porikli²,

Papanikolopoulos

2009

2009 IEEE Conference on Computer Vision and Pattern Recognition

491

234

View full text Add to dashboard Cite

One of the principal bottlenecks in applying learning techniques to classification problems is the large amount of labeled training data required. Especially for images and video, providing training data is very expensive in terms of human time and effort. In this paper we propose an active learning approach to tackle the problem. Instead of passively accepting random training examples, the active learning algorithm iteratively selects unlabeled examples for the user to label, so that human effort is focused on labeling the most "useful" examples. Our method relies on the idea of uncertainty sampling, in which the algorithm selects unlabeled examples that it finds hardest to classify. Specifically, we propose an uncertainty measure that generalizes margin-based uncertainty to the multi-class case and is easy to compute, so that active learning can handle a large number of classes and large data sizes efficiently. We demonstrate results for letter and digit recognition on datasets from the UCI repository, object recognition results on the Caltech-101 dataset, and scene categorization results on a dataset of 13 natural scene categories. The proposed method gives large reductions in the number of training examples required over random selection to achieve similar classification accuracy, with little computational overhead.

show abstract

Section: Previous Workmentioning

confidence: 99%

Multi-class active learning for image classification

Joshi

Porikli²,

Papanikolopoulos

2009

2009 IEEE Conference on Computer Vision and Pattern Recognition

491

234

View full text Add to dashboard Cite

show abstract

“…We compare the strangeness measure presented in our paper to two baseline methods which have been previously utilized for computing classifier confidence in active learning experiments [8,9]. For these, we utilize the boosting classifier evaluated in the previous section.…”

Section: Imagesmentioning

confidence: 99%

“…For this purpose, we utilize the strangeness measure from Eq. (8). When computing the strangeness for instances in G, we already know the semantic label for the instance and the strangeness computation is straightforward.…”

Section: Strangeness Measurementioning

confidence: 99%

“…Hence, a critical component is to provide informative unlabeled samples for the human to label and classification uncertainty estimates are employed for selecting such samples. A commonly used uncertainty measure is entropy [8] which is an information-theoretic criterion computed over label likelihoods for an instance. An alternative measure is the best versus second best heuristic [9] computed as the difference between the probability values of the two labels having the highest estimated probability value with a low difference value indicating more uncertainty.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Introspective semantic segmentation

Singh

Košecká

2014

IEEE Winter Conference on Applications of Computer Vision

View full text Add to dashboard Cite

Traditional approaches for semantic segmentation work in a supervised setting assuming a fixed number of semantic categories and require sufficiently large training sets. The performance of various approaches is often reported in terms of average per pixel class accuracy and global accuracy of the final labeling. When applying the learned models in the practical settings on large amounts of unlabeled data, possibly containing previously unseen categories, it is important to properly quantify their performance by measuring a classifier's introspective capability. We quantify the confidence of the region classifiers in the context of a non-parametric k-nearest neighbor (k-NN) framework for semantic segmentation by using the so called strangeness measure. The proposed measure is evaluated by introducing confidence based image ranking and showing its feasibility on a dataset containing a large number of previously unseen categories.

show abstract

“…In [26], the authors employ user inputs at multiple levels of granularity, also referred to as multi-level annotations. Holub et al [13] propose entropy-based active learning that can handle batch-mode selection in principle, however, the approach is prohibitively expensive in practice. Recently a few researchers have proposed batch-mode selection algorithms [3], [12], however these are restricted to only binary classification.…”

Section: Introductionmentioning

confidence: 99%

Multi-class batch-mode active learning for image classification

Joshiy

Porikli

Papanikolopoulos

2010

2010 IEEE International Conference on Robotics and Automation

View full text Add to dashboard Cite

Accurate image classification is crucial in many robotics and surveillance applications -for example, a vision system on a robot needs to accurately recognize the objects seen by its camera. Object recognition systems typically need a large amount of training data for satisfactory performance. The problem is particularly acute when many object categories are present. In this paper we present a batch-mode active learning framework for multi-class image classification systems. In active learning, images are to be chosen for interactive labeling, instead of passively accepting training data. Our framework addresses two important issues: i) it handles redundancy between images which is crucial when batch-mode selection is performed; and ii) we pose batch selection as a submodular function optimization problem that makes an inherently intractable problem efficient to solve, while having approximation guarantees. We show results on image classification data in which our approach substantially reduces the amount of training required over the baseline. ICRA 2010This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. Abstract-Accurate image classification is crucial in many robotics and surveillance applications -for example, a vision system on a robot needs to accurately recognize the objects seen by its camera. Object recognition systems typically need a large amount of training data for satisfactory performance. The problem is particularly acute when many object categories are present. In this paper we present a batch-mode active learning framework for multi-class image classification systems. In active learning, images are to be chosen for interactive labeling, instead of passively accepting training data. Our framework addresses two important issues: i) it handles redundancy between different images which is crucial when batch-mode selection is performed; and ii) we pose batchselection as a submodular function optimization problem that makes an inherently intractable problem efficient to solve, while having approximation guarantees. We show results on image classification data in which our approach substantially reduces the amount of training required over the baseline.

show abstract

Entropy-based active learning for object recognition

Cited by 215 publications

References 10 publications

Multi-class active learning for image classification

Multi-class active learning for image classification

Introspective semantic segmentation

Multi-class batch-mode active learning for image classification

Contact Info

Product

Resources

About