Open Cross-Domain Visual Search

Thong, William; Mettes, Pascal; Snoek, Cees G. M.

doi:10.48550/arxiv.1911.08621

Cited by 3 publications

(5 citation statements)

References 36 publications

(99 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model has utilized a generalization technique by reducing the MMD distance between several domains to ensure its ability to detect any abnormal event in an unobserved target domain. In addition, an open cross-domain visual search was created by [9] and implemented in a freehand sketch program. This refers to searching for pairs of target and source domains.…”

Section: Background and Literature Reviewsmentioning

confidence: 99%

Video-based Domain Generalization for Abnormal Event and Behavior Detection

Jarraya,

Almazroey

2024

IJACSA

View full text Add to dashboard Cite

Surveillance cameras have been widely deployed in public and private areas in recent years to enhance security and ensure public safety, necessitating the monitoring of unforeseen incidents and behaviors. An intelligent automated system is essential for detecting anomalies in video scenes to save the time and cost associated with manual detection by laborers monitoring displays. This study introduces a deep learning method to identify abnormal events and behaviors in surveillance footage of crowded areas, utilizing a scene-based domain generalization strategy. By utilizing the keyframe selection approach, keyframes containing relevant information are extracted from video frames. The chosen keyframes are utilized to create a spatio-temporal entropy template that reflects the motion area. The acquired template is then fed into the pre-trained AlexNet network to extract high-level features. The study utilizes the Relieff feature selection approach to choose suitable features, which are then served as input to Support Vector Machine (SVM) classifier. The model is assessed using six available datasets and two datasets built in this research, containing videos of normal and abnormal events and behaviors. The study found that the proposed method, utilizing domain generalization, surpassed state-of-arts methods in terms of detection accuracy, achieving a range from 87.5% to 100%. It also demonstrated the model's effectiveness in detecting anomalies from various domains with an accuracy rate of 97.13%.

show abstract

Section: Background and Literature Reviewsmentioning

confidence: 99%

Video-based Domain Generalization for Abnormal Event and Behavior Detection

Jarraya,

Almazroey

2024

IJACSA

View full text Add to dashboard Cite

show abstract

“…In recent years, motivated by the zero-shot validation criterion for supervised photo retrieval [168], zero-shot sketchbased image retrieval (ZS-SBIR) has been studied as a new 6. https://en.wikipedia.org/wiki/Topological sorting topic in free-hand sketch community [8], [169], [170], [171], [172], [173], [174], [175], [176], [177]. Similar to the natural photo zero-shot learning/recognition [178], [179], [180], the ZS-SBIR system aims to search the candidate photos for the sketch query that is from the unseen categories.…”

Section: Zero-shot Sbirmentioning

confidence: 99%

Deep Learning for Free-Hand Sketch: A Survey

Xu¹

2020

Preprint

View full text Add to dashboard Cite

Free-hand sketches are highly hieroglyphic and illustrative, which have been widely used by humans to depict objects or stories from ancient times to the present. The recent prevalence of touchscreen devices has made sketch creation a much easier task than ever and consequently made sketch-oriented applications increasingly more popular. The prosperity of deep learning has also immensely promoted the research for the free-hand sketch. This paper presents a comprehensive survey of the free-hand sketch oriented deep learning techniques. The main contents of this survey include: (i) The intrinsic traits and domain-unique challenges of the free-hand sketch are discussed, to clarify the essential differences between free-hand sketch and other data modalities, e.g., natural photo. (ii) The development of the free-hand sketch community in the deep learning era is reviewed, by surveying the existing datasets, research topics, and the state-of-the-art methods via a detailed taxonomy. (iii) Moreover, the bottlenecks, open problems, and potential research directions of this community have also been discussed to promote the future works.

show abstract

“…In the context of image retrieval, multiple works addressed the sketch-based image retrieval problem [52,8], even across multiple domains. In [40] the authors proposed a method to perform cross-domain image retrieval by training domainspecific experts. While these approaches integrated DA and ZSL, none of them considered the more complex scenario of DG, where no target data are available.…”

Section: Related Workmentioning

confidence: 99%

“…We build our validation + test set with 100 classes that contain at least 40 images per domain and that has no overlap with ImageNet. We reserve 45 of these classes for the unseen test set, matching the number used in [40], and the remaining 55 classes for the unseen validation set. The remaining 245 classes are used as seen classes during training.…”

Section: Datasets and Implementation Detailsmentioning

confidence: 99%

“…Finally, we report the final results on the 45 unseen classes. As semantic representation we use word2vec embeddings [27] extracted from the Google News corpus and L2 -normalized, following [40]. For all the baselines and our method, we employ as base architecture a ResNet-50 [13] pretrained on ImageNet, using the same number of epochs and SGD with momentum as optimizer, with the same hyperparameters of [40].…”

Section: Datasets and Implementation Detailsmentioning

confidence: 99%

See 1 more Smart Citation

Towards Recognizing Unseen Categories in Unseen Domains

Mancini¹,

Akata²,

Ricci³

et al. 2020

Preprint

View full text Add to dashboard Cite

Current deep visual recognition systems suffer from severe performance degradation when they encounter new images from classes and scenarios unseen during training. Hence, the core challenge of Zero-Shot Learning (ZSL) is to cope with the semantic-shift whereas the main challenge of Domain Adaptation and Domain Generalization (DG) is the domain-shift. While historically ZSL and DG tasks are tackled in isolation, this work develops with the ambitious goal of solving them jointly, i.e. by recognizing unseen visual concepts in unseen domains. We present CuMix (Curriculum Mixup for recognizing unseen categories in unseen domains), a holistic algorithm to tackle ZSL, DG and ZSL+DG. The key idea of CuMix is to simulate the test-time domain and semantic shift using images and features from unseen domains and categories generated by mixing up the multiple source domains and categories available during training. Moreover, a curriculum-based mixing policy is devised to generate increasingly complex training samples. Results on standard ZSL and DG datasets and on ZSL+DG using the DomainNet benchmark demonstrate the effectiveness of our approach.

show abstract

Open Cross-Domain Visual Search

Cited by 3 publications

References 36 publications

Video-based Domain Generalization for Abnormal Event and Behavior Detection

Video-based Domain Generalization for Abnormal Event and Behavior Detection

Deep Learning for Free-Hand Sketch: A Survey

Towards Recognizing Unseen Categories in Unseen Domains

Contact Info

Product

Resources

About