Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks
By Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, Yu Qiao. 2016
5,675 citations recorded by the source corpus.
Face detection and alignment in unconstrained environment are challenging due to various poses, illuminations, and occlusions. Recent studies show that deep learning approaches can achieve impressive performance on these two tasks. In this letter, we propose a deep cascaded multitask framework that exploits the inherent correlation between detection and alignment to boost up their performance. In particular, our framework leverages a cascaded architecture with three stages of carefully designed deep convolutional networks to predict face and landmark location in a coarse-to-fine manner. In addition, we propose a new online hard sample mining strategy that further improves the performance in practice. Our method achieves superior accuracy over the state-of-the-art techniques on the challenging face detection dataset and benchmark and WIDER FACE benchmarks for face detection, and annotated facial landmarks in the wild benchmark for face alignment, while keeps real-time performance.
Subfields: Computer Vision and Pattern Recognition, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Xiaogang Wang
- Ming-Hsuan Yang
- Andrew Zisserman
- Trevor Darrell
- Andrea Vedaldi
- Chen Change Loy
In Signal Processing
- Weihong Deng
- Stefanos P. Zafeiriou
- Ramalingam Chellappa
- Anil K. Jain
- Xiaoou Tang
- Weiyang Liu
- Jiankang Deng
- Xiaogang Wang
Citations in the corpus
- ScaIL: Classifier Weights Scaling for Class Incremental Learning (2020, methodology): “Face detection was done using MTCNN [33] which was applied to each image prior to training and test phases.”
- Compacting, Picking and Growing for Unforgetting Continual Learning (2019, methodology): “For these datasets, faces are aligned using MTCNN [50] with output size of 112 × 112.”
- SphereFace: Deep Hypersphere Embedding for Face Recognition (2017, methodology): “The face landmarks in all images are detected by MTCNN [39].”
- VGGFace2: A Dataset for Recognising Faces across Pose and Age (2017, methodology): “It includes human verified bounding boxes around faces, and five fiducial keypoints predicted by the model of [26].”
- A Comprehensive Study of Class Incremental Learning Algorithms for Visual Tasks (2020, methodology): “Face cropping is done with MTCNN [64] before further processing.”
- TVQA+: Spatio-Temporal Grounding for Video Question Answering (2019, methodology): “We also experimented with semi-automated annotation for people with face detection (Zhang et al., 2016) and recognition model (Liu et al.”
- Less-forgetful Learning for Domain Expansion in Deep Neural Networks (2017, background): “Deep neural networks (DNNs) have advanced to nearly human levels of object, face, and speech recognition (Taigman et al. 2014) (Graves, Mohamed, and Hinton 2013) (Szegedy et al. 2014) (Simonyan and Zisserman 2014) (Zhang et al. 2016) (Richardson, Reynolds, and Dehak 2015).”
- DeepFakes: a New Threat to Face Recognition? Assessment and Detection (2018, methodology): “Landmarks detection was done using publicly available pre-trained MTCNN model [22].”
- Deep Learning for Generic Object Detection: A Survey (2018, background): “ect detection, particularly for researchers wishing to enter the field. Since our focus is on generic object detection, the extensive work on DCNNs for specific object detection, such as face detection [154,306,116], pedestrian detection [307,109], vehicle detection [322] and traffic sign detection [32”
- CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition (2020, neutral): “We follow [8] to crop the 112 × 112 faces with five landmarks [37, 27].”
- Improved Knowledge Distillation for Training Fast Low Resolution Face Recognition Model (2019, methodology): “The faces in LFW are detected using MTCNN and aligned to 112×96 resolution.”
- Recent Advances in Deep Learning for Object Detection (2019, background): “[192] proposed a deep cas- 1564 caded multi-task face detector with cascaded structure (MTCNN).”
- Face identity disentanglement via latent space mapping (2020, methodology): “We then detect [Zhang et al. 2016] and crop faces from the output images and the real images, creating the test sets.”
- Complementary Incremental Hashing With Query-Adaptive Re-Ranking for Image Retrieval (2020, methodology): “Then the Multi-task CNN method (MTCNN) [46] is employed for face alignment.”
- Comparator Networks (2018, methodology): “As preprocessing, we predict 5 facial keypoints (Figure 3) over the entire dataset with a pre-trained MTCNN [33], and estimate three face poses by thresholding angle ratios.”