Deep learning algorithm with visual impression

Yang, Mengduo; Li, Fanzhang; Zhang, Li; Zhang, Zhao

doi:10.1016/j.ipl.2018.03.004

Cited by 9 publications

(13 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To ensure a good recall rate, anchors are thoroughly engineered based on the statistics computed from the training/validation set [79,80]. (b) Some design choices based on a particular dataset may not apply to other applications, which affects the generality [81]. (c) During the learning phase, the anchorbased approaches rely on intersection union (IoU) to define the positive/negative samples, thus adding extra computation and hyper-parameters for an object detection system [82].…”

Section: Anchor-based Detectorsmentioning

confidence: 99%

Object Detection Using Deep Learning, CNNs and Vision Transformers: A Review

Amjoud

Amrouch

2023

IEEE Access

View full text Add to dashboard Cite

Detecting objects remains one of computer vision and image understanding applications' most fundamental and challenging aspects. Significant advances in object detection have been achieved through improved object representation and the use of deep neural network models. This paper examines more closely how object detection has evolved in the era of deep learning over the past years. We present a literature review on various state-of-the-art object detection algorithms and the underlying concepts behind these methods. We classify these methods into three main groups: anchor-based, anchor-free, and transformer-based detectors. Those approaches are distinct in the way they identify objects in the image. We discuss the insights behind these algorithms and experimental analyses to compare quality metrics, speed/accuracy tradeoffs, and training methodologies. The survey compares the major convolutional neural networks for object detection. It also covers the strengths and limitations of each object detector model and draws significant conclusions. We provide simple graphical illustrations summarising the development of object detection methods under deep learning. Finally, we identify where future research will be conducted.

show abstract

Section: Anchor-based Detectorsmentioning

confidence: 99%

Object Detection Using Deep Learning, CNNs and Vision Transformers: A Review

Amjoud

Amrouch

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…They said that expressions, occultation or hindrances in the captured image, poses or face angle, illumination or lightness and darkness of the image, and facial features were the basic factors in face recognition. Deep learning shows ISSN: 2502-4752  excellent performance in face recognition [15]. However, it needs a large number of interpreted training datasets.…”

Section: Figure 1 Face Recognition Processmentioning

confidence: 99%

“…The system used a template-based approach to face recognition that compares images with sets of templates from a database [21]. Sets of templates were constructed using different algorithm tools like principal component analysis (PCA) [22], [23] using eigenfaces algorithm, linear discriminant analysis (LDA) [24] using fisher faces algorithm, support vector machine (SVM) [25] using local binary pattern histogram, the template matching algorithm [26], and the deep learning algorithm [15]. The researchers created the criteria evaluation of the system to identify the appropriate and best face recognition algorithm to be used in intruder detection.…”

Section: Algorithmsmentioning

confidence: 99%

Intruder detection and recognition using different image processing techniques for a proactive surveillance

Rodelas

Ballera

2021

IJEECS

View full text Add to dashboard Cite

To innovate a proactive surveillance camera, there is a need for efficient face detection and recognition algorithm. The researchers used one of the ViolaJones algorithm and used different image processing techniques to recognize intruders or not. The goal of the research is to recognize the fastest way on how the homeowners will be informed if an intruder or burglar enters their home using a proactive surveillance device. This device was programmed based on the different recognition algorithms and a criteria evaluation framework that could recognize intruders and burglars and the design used was developmental research to satisfy the research problem. The researchers used the Viola-Jones algorithm for face detection and five algorithms for face recognition. The criteria evaluation was used to identify the best face recognition algorithm and was tested in a real-world situation and captured a series of images camera and processed by proactive face detection and recognition. The result shows that the system can detect and recognize intruders and proactively send a notification to the homeowners via mobile application. It is concluded that the system can recognize the intruders and proactively notify the household members using the mobile applications and activate the alarm system of the house.

show abstract

“…In [6], Guided Anchoring, which uses semantic features to guide anchor setting, dynamically changes anchor shapes to suit different goals. In [7], an anchor generation function, which learns to dynamically generate anchors from custom prior boxes, is proposed. In [8], FSAF matches the most suitable FPN feature level for each real box by minimizing the training loss.…”

Section: Introductionmentioning

confidence: 99%

An anchor-free detector designed for dense small water object detection

Tao

Wang

2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

The detection of dense small objects on the water surface is one of the hot topics in object detection. In this paper, a one-to-many label assignment strategy based on the OTA algorithm, which is applied to anchor-free detector, is proposed to improve the detection accuracy of dense small objects on water surface. To be specific, one-to-many label assignment means that a ground truth (GT) corresponding to multiple prediction boxes is conducted globally. The cost function used by OTA algorithm is improved to make the distribution of positive and negative samples more reasonable. Meanwhile, an efficient training strategy is designed to accelerate network convergence by adding L1 loss in the last stage of training. The results show that the proposed strategies achieve 87.9% average precision (AP). Compared to the original algorithm, we achieve 3.3% relative improvement for the detection precision of dense small objects.

show abstract

Deep learning algorithm with visual impression

Cited by 9 publications

References 6 publications

Object Detection Using Deep Learning, CNNs and Vision Transformers: A Review

Object Detection Using Deep Learning, CNNs and Vision Transformers: A Review

Intruder detection and recognition using different image processing techniques for a proactive surveillance

An anchor-free detector designed for dense small water object detection

Contact Info

Product

Resources

About