Towards a deep learning framework for unconstrained face detection

Zheng, Yuejiu; Zhu, Chenchen; Luu, Khoa; Bhagavatula, Chandrasekhar; Le, Trung-Hieu; Savvides, Marios

doi:10.1109/btas.2016.7791203

Cited by 21 publications

(9 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the inconsistency in the bounding boxes in the AFLW dataset, we adopt the use of a face detector first to normalize the scale of the faces. The Multiple Scale Faster Region-based CNN approach [49] has shown good results and at a fast speed. We use the recent extension to this work, the Contextual Multi-Scale Region-based CNN (CMS-RCNN) approach [50] to perform the face detection in any experiment where face detection is needed.…”

Section: W-lpmentioning

confidence: 99%

Faster than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses

Bhagavatula

Zhu

Luu

et al. 2017

2017 IEEE International Conference on Computer Vision (ICCV)

Self Cite

118

View full text Add to dashboard Cite

Facial alignment involves finding a set of landmark points on an image with a known semantic meaning. However, this semantic meaning of landmark points is often lost in 2D approaches where landmarks are either moved to visible boundaries or ignored as the pose of the face changes. In order to extract consistent alignment points across large poses, the 3D structure of the face must be considered in the alignment step. However, extracting a 3D structure from a single 2D image usually requires alignment in the first place. We present our novel approach to simultaneously extract the 3D shape of the face and the semantically consistent 2D alignment through a 3D Spatial Transformer Network (3DSTN) to model both the camera projection matrix and the warping parameters of a 3D model. By utilizing a generic 3D model and a Thin Plate Spline (TPS) warping function, we are able to generate subject specific 3D shapes without the need for a large 3D shape basis. In addition, our proposed network can be trained in an end-to-end framework on entirely synthetic data from the 300W-LP dataset. Unlike other 3D methods, our approach only requires one pass through the network resulting in a faster than realtime alignment. Evaluations of our model on the Annotated Facial Landmarks in the Wild (AFLW) and AFLW2000-3D datasets show our method achieves state-of-the-art performance over other 3D approaches to alignment.

show abstract

Section: W-lpmentioning

confidence: 99%

Faster than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses

Bhagavatula

Zhu

Luu

et al. 2017

2017 IEEE International Conference on Computer Vision (ICCV)

Self Cite

118

View full text Add to dashboard Cite

show abstract

“…When it comes to accuracy, as the experimental result is presented in Fig. 4, YOLOFace has achieved the best performance with an accuracy as high as 98.96% which proves the effectiveness of CNN architectures for face detection Chi et al, 2017;Li et al, 2015;Mehta et al, 2018;Zheng et al, 2016). As it is mentioned in Introduction, face detection mechanism has already been integrated into the digital cameras and smartphones as well.…”

Section: Resultsmentioning

confidence: 69%

An Experimental Performance Comparison of Widely Used Face Detection Tools

Kabakuş

2019

ADCAIJ

View full text Add to dashboard Cite

Face detection is the task of detecting faces on photos, videos as well as the streaming data such as a webcam. Face detection, which is a specific type of general-purpose object detection, is a key prerequisite for many other artificial intelligence tasks such as face verification, face tagging and retrieval, and face tracking. In addition to that, nowadays, face detection is commonly used in daily routines such as social media, and camera software of smartphones. As a result of this necessity, several face detection tools have been proposed. In this study, an experimental performance comparison of well-known face detection tools in terms of (1) accuracy, and (2) elapsed time of detection, which has become even more critical criteria especially when the face detection mechanism is utilized for a real-time system, is proposed. As a result of this experimental study, it is aimed that shed light on the much-concerned query “which face detection tool provides the best performance?”. In addition to that, this study succeeds in showing that convolutional neural networks achieve great accuracy for face detection.

show abstract

“…These methods are not very effective in feature extraction and also contain tedious manual feature extraction steps. Meanwhile, due to the continuous optimization of Graphic Processing Unit (GPU) and network architectures, deep learning methods which can substantially improve recognition results have gradually become mature and perfect its wider application [27,28].…”

Section: Related Workmentioning

confidence: 99%

Generation of Head Mirror Behavior and Facial Expression for Humanoid Robots

Liu¹,

Li²,

Zhang³

et al. 2021

AAIML

View full text Add to dashboard Cite

Understanding of human expressions and action intentions is crucial to improve the intelligence of the social robots. The social robots can also perform mirror behaviors to learn from and understand others. Here we present an intelligent framework in which a humanoid social robot can simulate human facial expressions and emulate human head motions. To fulfill these requirements, an intelligent interactive robot control framework which can simulate human facial expressions and head motions is proposed. We developed a physical animatronic robotic head platform with soft skin, whose expressions can be adjusted through customized mechanical dynamic structure. A vision-based facial expression and head motion recognition method is proposed to establish the understanding mechanism of the expression and action intention of the intelligent robot using deep learning framework. Finally, the recognition results are employed to determine the optimal motor displacements for transferring the head states of a human to the robot. The experimental results demonstrate that our method can maintain high accuracy when facing different human subjects. The comprehensive evaluation shows that our method can perform accurate and real-time facial expression generation and head mirror behavior, which has a promising application for home assistant robots.

show abstract

Towards a deep learning framework for unconstrained face detection

Cited by 21 publications

References 22 publications

Faster than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses

Faster than Real-Time Facial Alignment: A 3D Spatial Transformer Network Approach in Unconstrained Poses

An Experimental Performance Comparison of Widely Used Face Detection Tools

Generation of Head Mirror Behavior and Facial Expression for Humanoid Robots

Contact Info

Product

Resources

About