Fabric Retrieval Based on Multi-Task Learning

Xiang, Jun; Zhang, Ning; Pan, Ruru; Gao, Weidong

doi:10.1109/tip.2020.3043877

Cited by 23 publications

(29 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the weakness was that it produced a high error rate for images with many patterns (i.e., printed fabrics). Multi-task learning (MTL) combines a modified CNN with deep hash coding to retrieve fabric images [33]. MTL aims to improve the prediction accuracy and learning efficiency of each task in comparison with training a separate model.…”

Section: Feature Extraction Based On Cnn Methodsmentioning

confidence: 99%

“…Another benefit is that they allow for image retrieval on large datasets and use a deep layer to obtain more specific feature parameters, which are then used in similarity matching. In general, the features of fully connected CNN layers are used to match query and database images [2], [33], [45]. In several studies, the features extracted in the convolutional layer were used for image retrieval [31], [46].…”

Section: Feature Extraction Based On Cnn Methodsmentioning

confidence: 99%

“…Future research developments are suggested to modify CNN methods, which comprises global features capable of integrating with local feature descriptors, such as images with intricate texture patterns [39]. Integrating CNNs and hash encoding aims to reduce feature dimensions and computation time [2], [33], [43], [47]. Furthermore, its features are strengthened by the color filter process to produce color features that are more resistant to illumination and background [35].…”

Section: Comparison Of Traditional and Cnn Feature Extrationmentioning

confidence: 99%

“…Tests should also be carried out involving wrinkling fabric to capture pictures of material with different degrees of creasing. For example, hash coding can be used to reduce feature dimensions of CNN output and computation time [2], [33], [49]. Pruning neural networks can be done to reduce the computational load on the CNN method.…”

Section: Comparison Of Traditional and Cnn Feature Extrationmentioning

confidence: 99%

See 3 more Smart Citations

Content-based image retrieval for fabric images: A survey

Tena

Hartanto

Ardiyanto

2021

IJEECS

View full text Add to dashboard Cite

In <span>recent years, a great deal of research has been conducted in the area of fabric image retrieval, especially the identification and classification of visual features. One of the challenges associated with the domain of content-based image retrieval (CBIR) is the semantic gap between low-level visual features and high-level human perceptions. Generally, CBIR includes two main components, namely feature extraction and similarity measurement. Therefore, this research aims to determine the content-based image retrieval for fabric using feature extraction techniques grouped into traditional methods and convolutional neural networks (CNN). Traditional descriptors deal with low-level features, while CNN addresses the high-level, called semantic features. Traditional descriptors have the advantage of shorter computation time and reduced system requirements. Meanwhile, CNN descriptors, which handle high-level features tailored to human perceptions, deal with large amounts of data and require a great deal of computation time. In general, the features of a CNN's fully connected layers are used for matching query and database images. In several studies, the extracted features of the CNN's convolutional layer were used for image retrieval. At the end of the CNN layer, hash codes are added to reduce </span>search time.

show abstract

Section: Feature Extraction Based On Cnn Methodsmentioning

confidence: 99%

Section: Feature Extraction Based On Cnn Methodsmentioning

confidence: 99%

Section: Comparison Of Traditional and Cnn Feature Extrationmentioning

confidence: 99%

Section: Comparison Of Traditional and Cnn Feature Extrationmentioning

confidence: 99%

See 2 more Smart Citations

Content-based image retrieval for fabric images: A survey

Tena

Hartanto

Ardiyanto

2021

IJEECS

View full text Add to dashboard Cite

show abstract

“…The foreground information is sufficient for computing, identifying, and searching through the pictures and flow chart for feature calculation shown in Figure 8. The following algorithm has been proposed to do this [19].…”

Section: Algorithmmentioning

confidence: 99%

Design a Framework for Content Based Image Retrieval Using Hybrid Features Analysis

Kumar¹,

Singh²,

Raja³

et al. 2021

View full text Add to dashboard Cite

In recent years, scholars have found content-based image recovery to be a particularly interesting and exciting field. This field develops quick image recovery algorithms that are very similar to pictures from various data sources. There is currently a lot of space and sources for data storage. Finally, once data sources are obtained, there is just competition for the exact and best result. The result is filtered using the heuristic correlation based on the above features of the photographs. The overall matching scores are calculated by adding all these individual feature ratings. The recommended method will be used to retrieve all images containing the content of the query image. The scores will be used to rate the match. The results of the combined method simulation demonstrate that the strategy is successful. This proposed model will improve the accuracy of search results. This research model, like the other metadata models on the web, is interactive and familiar for searching pictures from huge databases and data sources. This model also has a number of handy features built in to improve precision and efficiency. Although there are a few little mysteries in this model, they must be answered subsequently.

show abstract