Single-Shot Scale-Aware Network for Real-Time Face Detection

Zhang, Shifeng; Wen, Longyin; Shi, Hailin; Lei, Zhen; Lyu, Siwei; Li, Stan Z.

doi:10.1007/s11263-019-01159-3

Cited by 52 publications

(28 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…OverFeat [31] is one of the first one-stage detectors and since then, several other methods have been proposed, such as YOLO [26,27] and SSD [24]. Recent researches on one-stage approach focus on enriching features for detection [8], designing different architecture [39] and addressing class imbalance issue [41,22,40]. Train-from-scratch object detectors.…”

Section: Related Workmentioning

confidence: 99%

ScratchDet: Training Single-Shot Object Detectors From Scratch

Zhu

Zhang

Wang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Self Cite

138

View full text Add to dashboard Cite

Current state-of-the-art object objectors are fine-tuned from the off-the-shelf networks pretrained on large-scale classification dataset ImageNet, which incurs some additional problems: 1) The classification and detection have different degrees of sensitivity to translation, resulting in the learning objective bias; 2) The architecture is limited by the classification network, leading to the inconvenience of modification. To cope with these problems, training detectors from scratch is a feasible solution. However, the detectors trained from scratch generally perform worse than the pretrained ones, even suffer from the convergence issue in training. In this paper, we explore to train object detectors from scratch robustly. By analysing the previous work on optimization landscape, we find that one of the overlooked points in current trained-from-scratch detector is the BatchNorm. Resorting to the stable and predictable gradient brought by BatchNorm, detectors can be trained from scratch stably while keeping the favourable performance independent to the network architecture. Taking this advantage, we are able to explore various types of networks for object detection, without suffering from the poor convergence. By extensive experiments and analyses on downsampling factor, we propose the Root-ResNet backbone network, which makes full use of the information from original images. Our ScratchDet achieves the state-of-the-art accuracy on PASCAL VOC 2007, 2012 and MS COCO among all the train-from-scratch detectors and even performs better than several one-stage pretrained methods. Codes will be made publicly available at https: //github.com

show abstract

Section: Related Workmentioning

confidence: 99%

ScratchDet: Training Single-Shot Object Detectors From Scratch

Zhu

Zhang

Wang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

Self Cite

138

View full text Add to dashboard Cite

show abstract

“…The use of faces is important for many applications including surveillance systems, human-machine interaction, and airports [1]. Considering the face is a unique identification for each human being, in addition to that humans are the most causing of insecurity in the society, the summarization via faces detection and then identification can be useful to identify the bad behaviors person in the stadium for example [2,3].…”

Section: Acknowledgmentmentioning

confidence: 99%

Face Recognition and Summarization for Surveillance Video Sequences

Almaadeed¹,

Elharrouss²

2020

University of the Future: Re-Imagining Research and Higher Education

View full text Add to dashboard Cite

Face recognition and video summarization represent challenging tasks for several computer vision applications including video surveillance, criminal investigations, and sports applications. For long videos, it is difficult to search within a video for a specific action and/or person. Usually, human action recognition approaches presented in the literature deal with videos that contain only a single person, and they are able to recognize his action. This paper proposes an effective approach to multiple human action detection, recognition, and summarization. The multiple action detection extracts human bodies’ silhouette then generates a specific sequence for each one of them using motion detection and tracking method. Each of the extracted sequences is then divided into shots that represent homogeneous actions in the sequence using the similarity between each pair frames. Using the histogram of the oriented gradient (HOG) of the temporal difference map (TDMap) of the frames of each shot, we recognize the action by performing a comparison between the generated HOG and the existed HOGs in the training phase which represents all the HOGs of many actions using a set of videos for training.

show abstract

“…Every loss function from the training sample is added IoU-aware weight to increase accuracy. The SFDet method is able to recognize faces in real-time [7].…”

Section: Related Workmentioning

confidence: 99%

Real-Time Face Recognition with Eigenface Method

Wirdiani¹,

Lattifia²,

Supadma³

et al. 2019

IJIGSP

View full text Add to dashboard Cite

Real-time face image recognition is a face recognition system that is done directly using a webcam camera from a computer. Face recognition system aims to implement a biometrics system as a real-time facial recognition system. This system is divided into two important processes, namely the training process and the identification process. The registration process is a process where a user registered their name in a system and then registers their face. Face data that has been registered will be used for the next process, namely the identification process. The face registration process uses face detection using the OpenCV library. The feature extraction process and introduction to the recognition system use the Eigenface method. The results of this study found that, the Eigenface method is able to detect faces accurately up to 4 people simultaneously. The greater the threshold value will result in a greater value of FRR, while there isn't any FAR value found from different thresholds. The level of lighting, poses, and facial distance from the camera when training and testing the face image heavily influences the use of the eigenface method.

show abstract

Single-Shot Scale-Aware Network for Real-Time Face Detection

Cited by 52 publications

References 64 publications

ScratchDet: Training Single-Shot Object Detectors From Scratch

ScratchDet: Training Single-Shot Object Detectors From Scratch

Face Recognition and Summarization for Surveillance Video Sequences

Real-Time Face Recognition with Eigenface Method

Contact Info

Product

Resources

About