Survey of Video Based Small Target Detection

Liu, Ying; Geng, Luyao; Zhang, Weidong; Yan, Gao; Xu, Zhijie

doi:10.18178/joig.9.4.122-134

Cited by 82 publications

(16 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Where FPR and FNR were the false positive rate and false negative rate calculated in leave-one-subjectout (LOSO) cross test [19,24]. The LOSO cross test was to leave data samples of one subject in the training set as the test samples, and the data samples of the remained subjects as the training samples.…”

Section: Methodsmentioning

confidence: 99%

The Effect of Active break on Autonomic Nervous Pattern Recognition of Learning States

Sun

zhang

Wang

et al. 2022

Preprint

View full text Add to dashboard Cite

Purpose: This paper explores the effect of active break on the distinguishability of autonomic nervous patterns of learning states through machine learning. Method: First, we collected electrocardiogram (ECG) data of 77 subjects before and after active break, and accurately located the R-wave peaks from the ECG signal to calculate the RR interval series. Second, the RR interval samples were segmented according to the inclusion criteria of certain learning states. The initial 39 ECG features were empirically calculated, and the optimal feature combination for learning states recognition was selected through sequential backward selection and leave-one-subject-out cross test. Finally, we established binary-classification models of pairs of learning states and compared their performance of learning states recognition before and after active break. Discussion: Active break can promote or inhibit the students’ academic performance. Besides statistical analysis shows the stability of the cognitive ability. What’s more active break increases the physiological response to fatigue Conclusion: (1) The autonomic nervous patterns of knowledge input-processing and retrieval-processing and those of cognitive load matching and mismatching in knowledge retrieval processing became less distinguishable with the effect of active break. (2) The autonomic nervous patterns of mental fatigue and no fatigue states became more distinguishable with the effect of active break. (3) Stronger parasympathetic nervous activities make students achieve better academic performance during using new knowledge to solve problem stage.

show abstract

Section: Methodsmentioning

confidence: 99%

The Effect of Active break on Autonomic Nervous Pattern Recognition of Learning States

Sun

zhang

Wang

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Here, f, e, and l respectively denote the foreign body, the power equipment, and the label as shown in Figure 1. () 3 adds an extra edge from the original input to the output node if they are at the same level several to optimize the cross-scale connections, which allows scale-wise level reweighting to fuse more features without adding much cost [20]. Therefore, we use BiFPN layers to replace the FPN + PAN layers.…”

Section: Iccee-2022mentioning

confidence: 99%

Transmission Line Foreign Body Fault Detection Using Multi-Feature Fusion Based on Modified YOLOv5

Wang

et al. 2022

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

The particular work environment of transmission lines makes it liable to malfunction because of the foreign bodies attached to the power equipment. However, traditional detection methods are difficult to identify the exact category of foreign bodies and ignore the relationship between a foreign body and power equipment, which is correlated to determining whether there is a fault or not. In order to detect and distinguish the foreign bodies and give fault warnings automatically in transmission line patrol with high accuracy and efficiency, we improve the original YOLOv5 object detection model and propose a novel fusion detection model using multi-scale appearance and relationship features. The proposed model is capable of transmission line foreign body category detection and fault determination. By comparing with the original YOLOv5, the proposed model shows better accuracy and achieves the strong ability to give the fault determination simultaneously, which makes it convenient for power system operators and protects the safe and stable operation of the power system.

show abstract

“…In the early stage of video understanding, researchers pay more attention to the features of manual design, which is the basis for encoding the appearance and motion information of video 16 . With the great success of deep neural network in Imagenet 17,18 and object recognition and detection 43,44,45 , many video recognition and classification methods begin to extract features from 2D image convolution network after video frame extraction, and explore the improvement of video classification effect in combination with video optical flow information 20,19,21 . Such methods often process RGB image frames and optical flow image frames respectively, and fuse the features before recognition.…”

Section: Related Workmentioning

confidence: 99%

Improved video classification method based on non-parametric attention combined with self-supervision

L²

2022

Fourteenth International Conference on Digital Image Processing (ICDIP 2022)

View full text Add to dashboard Cite

It is worth mentioning that in the video sequence modeling, the best recognition architecture is transformer. The current popular transformer based video classification methods focus on the importance of current features in time sequence. The degree of characterization of simultaneous order is insufficient, and simple data augmentation has unstable classification effect. In this paper we proposed a method of non-parametric attention combined with self-supervised feature construction to further improve video classification. In this method, the non-parametric attention mechanism is constructed in the simultaneous order feature to fit the multi-local extreme value distribution. At the same time, in the process of model learning, the input video is randomly masked in temporal domain and spatial domain, and selfsupervised information is added to effectively learn the details and classification information of video content. Experiments using kinetics400, kinetics600 and something V2 datasets show that the algorithm in this paper has better improvement in accuracy than the current optimal method.

show abstract

Survey of Video Based Small Target Detection

Cited by 82 publications

References 0 publications

The Effect of Active break on Autonomic Nervous Pattern Recognition of Learning States

The Effect of Active break on Autonomic Nervous Pattern Recognition of Learning States

Transmission Line Foreign Body Fault Detection Using Multi-Feature Fusion Based on Modified YOLOv5

Improved video classification method based on non-parametric attention combined with self-supervision

Contact Info

Product

Resources

About