Harnessing noisy Web images for deep representation

Vo, Phong; Gînsca, Alexandru-Lucian; Borgne, Hervé Le; Popescu, Adrian

doi:10.1016/j.cviu.2017.01.009

Cited by 9 publications

(9 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To address this drawback, learning with noisy labels has been extensively studied in the AI related tasks. Recent works (Vo et al 2017;Chen and Gupta 2015) have achieved superior results in their tasks by using web data. Chen et al (Chen, Shrivastava, and Gupta 2013) use a semi-supervised learning algorithm to find the relationships between common sense and labeled images of given categories.…”

Section: Related Work Deep Learning From Noisy Web Datamentioning

confidence: 99%

Learning from Web Data Using Adversarial Discriminative Neural Networks for Fine-Grained Classification

Sun

Chen

Yang

2019

AAAI

View full text Add to dashboard Cite

Fine-grained classification is absorbed in recognizing the subordinate categories of one field, which need a large number of labeled images, while it is expensive to label these images. Utilizing web data has been an attractive option to meet the demands of training data for convolutional neural networks (CNNs), especially when the well-labeled data is not enough. However, directly training on such easily obtained images often leads to unsatisfactory performance due to factors such as noisy labels. This has been conventionally addressed by reducing the noise level of web data. In this paper, we take a fundamentally different view and propose an adversarial discriminative loss to advocate representation coherence between standard and web data. This is further encapsulated in a simple, scalable and end-to-end trainable multi-task learning framework. We experiment on three public datasets using large-scale web data to evaluate the effectiveness and generalizability of the proposed approach. Extensive experiments demonstrate that our approach performs favorably against the state-of-the-art methods.

show abstract

Section: Related Work Deep Learning From Noisy Web Datamentioning

confidence: 99%

Learning from Web Data Using Adversarial Discriminative Neural Networks for Fine-Grained Classification

Sun

Chen

Yang

2019

AAAI

View full text Add to dashboard Cite

show abstract

“…In this area, impressive improvement has been observed [6], [11], [12], [14]- [17]. For web data collections, Chen et al [14] use a semi-supervised learning algorithm to find the relationships between common sense and labeled images of given categories.…”

Section: Related Work a Deep Learning From Web Datamentioning

confidence: 99%

“…Year # imgs Source Flickr-CIFAR [15] 2014 230173 Flickr Flickr [6] 2015 1.2M Flickr Google [6] 2015 1.5M Google YFCC100M [19] 2015 99.3M Flickr Clothing [17] 2015 100000 Shopping sites Sketch [33] 2016 191067 Google Openimages [34] 2016 9.01M Google M-Flower-620 [35] 2016 20211 Instagram YouTube-8M [36] 2016 1.9B YouTube Weakly (Bird) [37] 2016 200000 Flickr Goldfince [11] 2016 9.8M Google Flickr-Bing 100 [12] [11] collect datasets from Google Images and conduct two rounds of cleaning: active learning and human in the loop. Xiao et al [17] establish a clothing dataset by searching images from shopping websites, which has both noisy labels and clean labels obtained by manual refinement.…”

Section: Datasetmentioning

confidence: 99%

“…Web data collection is related to the technique of image retrieval [38]- [41], but the existing engines are mainly based on traditional keyword search techniques. Similar collection methods are also used in works [12], [19], [35], [36], [42] which utilize web data for specific tasks.…”

Section: Datasetmentioning

confidence: 99%

“…In response to the inaccurate tags, Krause et al [11] remove images that appear in search results for more than one category to reduce the effect of noise. Vo et al [12] filter images by ranking the predicted results of classifiers trained on features exacted from a CNN. Nevertheless, such methods do not consider the complex relationships between the tag and the content of web images.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Recognition From Web Data: A Progressive Filtering Approach

Yang

Sun

Lai

et al. 2018

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Leveraging the abundant number of web data is a promising strategy in addressing the problem of data lacking when training convolutional neural networks (CNNs). However, the web images often contain incorrect tags, which may compromise the learned CNN model. To address this problem, this paper focuses on image classification and proposes to iterate between filtering out noisy web labels and fine-tuning the CNN model using the crawled web images. Overall, the proposed method benefits from the growing modeling capability of the learned model to correct labels for web images and learning from such new data to produce a more effective model. Our contribution is two-fold. First, we propose an iterative method that progressively improves the discriminative ability of CNNs and the accuracy of web image selection. This method is beneficial toward selecting high-quality web training images and expanding the training set as the model gets ameliorated. Second, since web images are usually complex and may not be accurately described by a single tag, we propose to assign a web image multiple labels to reduce the impact of hard label assignment. This labeling strategy mines more training samples to improve the CNN model. In the experiments, we crawl 0.5 million web images covering all categories of four public image classification data sets. Compared with the baseline which has no web images for training, we show that the proposed method brings notable improvement. We also report the competitive recognition accuracy compared with the state of the art.

show abstract