1
Joint Face Detection and Alignment using
Multi-task Cascaded Convolutional Networks
Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, Senior Member, IEEE, and Yu Qiao, Senior Member, IEEE
Abstract—Face detection and alignment in unconstrained en-
vironment 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
paper, we propose a deep cascaded multi-task framework which
exploits the inherent correlation between them to boost up their
performance. In particular, our framework adopts a cascaded
structure with three stages of carefully designed deep convolu-
tional networks that predict face and landmark location in a
coarse-to-fine manner. In addition, in the learning process, we
propose a new online hard sample mining strategy that can im-
prove the performance automatically without manual sample
selection. Our method achieves superior accuracy over the
state-of-the-art techniques on the challenging FDDB and WIDER
FACE benchmark for face detection, and AFLW benchmark for
face alignment, while keeps real time performance.
Index Terms—Face detection, face alignment, cascaded con-
volutional neural network
I. INTRODUCTION
F
ACE detection and alignment are essential to many face
applications, such as face recognition and facial expression
analysis. However, the large visual variations of faces, such as
occlusions, large pose variations and extreme lightings, impose
great challenges for these tasks in real world applications.
The cascade face detector proposed by Viola and Jones [2]
utilizes Haar-Like features and AdaBoost to train cascaded
classifiers, which achieve good performance with real-time
efficiency. However, quite a few works [1, 3, 4] indicate that
this detector may degrade significantly in real-world applica-
tions with larger visual variations of human faces even with
more advanced features and classifiers. Besides the cascade
structure, [5, 6, 7] introduce deformable part models (DPM) for
face detection and achieve remarkable performance. However,
they need high computational expense and may usually require
expensive annotation in the training stage. Recently, convolu-
tional neural networks (CNNs) achieve remarkable progresses
in a variety of computer vision tasks, such as image classifica-
tion [9] and face recognition [10]. Inspired by the good per-
K.-P. Zhang, Z.-F. Li and Y. Q. are with the Multimedia Laboratory,
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences,
Shenzhen 518055, China. E-mail: [email protected]; [email protected];
Z.-P. Zhang is with the Department of Information Engineering, The Chi-
nese University of Hong Kong, Hong Kong. E-mail: [email protected]
formance of CNNs in computer vision tasks, some of the CNNs
based face detection approaches have been proposed in recent
years. Yang et al. [11] train deep convolution neural networks
for facial attribute recognition to obtain high response in face
regions which further yield candidate windows of faces.
However, due to its complex CNN structure, this approach is
time costly in practice. Li et al. [19] use cascaded CNNs for
face detection, but it requires bounding box calibration from
face detection with extra computational expense and ignores
the inherent correlation between facial landmarks localization
and bounding box regression.
Face alignment also attracts extensive interests. Regres-
sion-based methods [12, 13, 16] and template fitting ap-
proaches [14, 15, 7] are two popular categories. Recently,
Zhang et al. [22] proposed to use facial attribute recognition as
an auxiliary task to enhance face alignment performance using
deep convolutional neural network.
However, most of the available face detection and face
alignment methods ignore the inherent correlation between
these two tasks. Though there exist several works attempt to
jointly solve them, there are still limitations in these works. For
example, Chen et al. [18] jointly conduct alignment and detec-
tion with random forest using features of pixel value difference.
But, the handcraft features used limits its performance. Zhang
et al. [20] use multi-task CNN to improve the accuracy of
multi-view face detection, but the detection accuracy is limited
by the initial detection windows produced by a weak face de-
tector.
On the other hand, in the training process, mining hard
samples in training is critical to strengthen the power of de-
tector. However, traditional hard sample mining usually per-
forms an offline manner, which significantly increases the
manual operations. It is desirable to design an online hard
sample mining method for face detection and alignment, which
is adaptive to the current training process automatically.
In this paper, we propose a new framework to integrate these
two tasks using unified cascaded CNNs by multi-task learning.
The proposed CNNs consist of three stages. In the first stage, it
produces candidate windows quickly through a shallow CNN.
Then, it refines the windows to reject a large number of
Advertisement
non-faces windows through a more complex CNN. Finally, it
uses a more powerful CNN to refine the result and output facial
landmarks positions. Thanks to this multi-task learning
framework, the performance of the algorithm can be notably
improved. The major contributions of this paper are summa-
rized as follows: (1) We propose a new cascaded CNNs based
framework for joint face detection and alignment, and carefully
2
TABLE I
COMPARISON OF SPEED AND VALIDATION ACCURACY OF OUR CNNS AND
PREVIOUS CNNS [19]
Group
CNN
300 Times Forward
Accuracy
Group1
Group1
Group2
Group2
Group3
Group3
12-Net [19]
P-Net
24-Net [19]
R-Net
48-Net [19]
O-Net
0.038s
0.031s
0.738s
0.458s
3.577s
1.347s
94.4%
94.6%
95.1%
95.4%
93.2%
95.4%
Stage 3: This stage is similar to the second stage, but in this
stage we aim to describe the face in more details. In particular,
the network will output five facial landmarks’ positions.
B. CNN Architectures
In [19], multiple CNNs have been designed for face detec-
tion. However, we noticed its performance might be limited by
the following facts: (1) Some filters lack diversity of weights
that may limit them to produce discriminative description. (2)
Compared to other multi-class objection detection and classi-
fication tasks, face detection is a challenge binary classification
task, so it may need less numbers of filters but more discrimi-
nation of them. To this end, we reduce the number of filters and
change the 5×5 filter to a 3×3 filter to reduce the computing
while increase the depth to get better performance. With these
improvements, compared to the previous architecture in [19],
we can get better performance with less runtime (the result is
shown in Table 1. For fair comparison, we use the same data for
both methods). Our CNN architectures are showed in Fig. 2.
C. Training
We leverage three tasks to train our CNN detectors:
face/non-face classification, bounding box regression, and
facial landmark localization.
1) Face classification: The learning objective is formulated as
a two-class classification problem. For each sample , we use
the cross-entropy loss:
(1)
where is the probability produced by the network that indi-
denotes
cates a sample being a face. The notation
the ground-truth label.
2) Bounding box regression: For each candidate window, we
predict the offset between it and the nearest ground truth (i.e.,
the bounding boxes’ left top, height, and width). The learning
objective is formulated as a regression problem, and we employ
the Euclidean loss for each sample :
‖ ̂
‖
(2)
where ̂
regression target obtained from the network and
is the ground-truth coordinate. There are four coordinates,
including left top, height and width, and thus
3) Facial landmark localization: Similar to the bounding box
regression task, facial landmark detection is formulated as a
.
Fig. 1. Pipeline of our cascaded framework that includes three-stage mul-
ti-task deep convolutional networks. Firstly, candidate windows are produced
through a fast Proposal Network (P-Net). After that, we refine these candidates
in the next stage through a Refinement Network (R-Net). In the third stage,
The Output Network (O-Net) produces final bounding box and facial land-
Advertisement
marks position.
design lightweight CNN architecture for real time performance.
(2) We propose an effective method to conduct online hard
sample mining to improve the performance. (3) Extensive ex-
periments are conducted on challenging benchmarks, to show
the significant performance improvement of the proposed ap-
proach compared to the state-of-the-art techniques in both face
detection and face alignment tasks.
II. APPROACH
In this section, we will describe our approach towards joint
face detection and alignment.
A. Overall Framework
The overall pipeline of our approach is shown in Fig. 1.
Given an image, we initially resize it to different scales to build
an image pyramid, which is the input of the following
three-stage cascaded framework:
Stage 1: We exploit a fully convolutional network[?], called
Proposal Network (P-Net), to obtain the candidate windows
and their bounding box regression vectors in a similar manner
as [29]. Then we use the estimated bounding box regression
vectors to calibrate the candidates. After that, we employ
non-maximum suppression (NMS) to merge highly overlapped
candidates.
Stage 2: all candidates are fed to another CNN, called Refine
Network (R-Net), which further rejects a large number of false
candidates, performs calibration with bounding box regression,
and NMS candidate merge.
3
Fig. 2. The architectures of P-Net, R-Net, and O-Net, where “MP” means max pooling and “Conv” means convolution. The step size in convolution and pooling
is 1 and 2, respectively.
regression problem and we minimize the Euclidean loss:
‖ ̂
‖
(3)
is the facial landmark’s coordinate obtained
where ̂
is the ground-truth coordinate.
from the network and
There are five facial landmarks, including left eye, right eye,
nose, left mouth corner, and right mouth corner, and thus
.
4) Multi-source training: Since we employ different tasks in
each CNNs, there are different types of training images in the
learning process, such as face, non-face and partially aligned
face. In this case, some of the loss functions (i.e., Eq. (1)-(3) )
are not used. For example, for the sample of background region,
, and the other two losses are set as 0.
we only compute
This can be implemented directly with a sample type indicator.
Then the overall learning target can be formulated as:
∑ ∑
(4)
where is the number of training samples. denotes on the
task importance. We use
in P-Net
and R-Net, while
in O-Net for more accurate facial land-
is the sample type indicator. In
marks localization.
this case, it is natural to employ stochastic gradient descent to
train the CNNs.
5) Online Hard sample mining: Different from conducting
traditional hard sample mining after original classifier had been
trained, we do online hard sample mining in face classification
task to be adaptive to the training process.
In particular, in each mini-batch, we sort the loss computed
in the forward propagation phase from all samples and select
the top 70% of them as hard samples. Then we only compute
the gradient from the hard samples in the backward propagation
phase. That means we ignore the easy samples that are less
helpful to strengthen the detector while training. Experiments
show that this strategy yields better performance without
manual sample selection. Its effectiveness is demonstrated in
the Section III.
III. EXPERIMENTS
In this section, we first evaluate the effectiveness of the
proposed hard sample mining strategy. Then we compare our
face detector and alignment against the state-of-the-art methods
in Face Detection Data Set and Benchmark (FDDB) [25],
WIDER FACE [24], and Annotated Facial Landmarks in the
Wild (AFLW) benchmark [8]. FDDB dataset contains the an-
notations for 5,171 faces in a set of 2,845 images. WIDER
FACE dataset consists of 393,703 labeled face bounding boxes
in 32,203 images where 50% of them for testing into three
subsets according to the difficulty of images, 40% for training
and the remaining for validation. AFLW contains the facial
landmarks annotations for 24,386 faces and we use the same
test subset as [22]. Finally, we evaluate the computational ef-
ficiency of our face detector.
Advertisement
A. Training Data
the
that
(i) Negatives: Regions
Since we jointly perform face detection and alignment, here
we use four different kinds of data annotation in our training
process:
Intersec-
tion-over-Union (IoU) ratio less than 0.3 to any ground-truth
faces; (ii) Positives: IoU above 0.65 to a ground truth face; (iii)
Part faces: IoU between 0.4 and 0.65 to a ground truth face; and
(iv) Landmark faces: faces labeled 5 landmarks’ positions.
Negatives and positives are used for face classification tasks,
positives and part faces are used for bounding box regression,
and landmark faces are used for facial landmark localization.
The training data for each network is described as follows:
1) P-Net: We randomly crop several patches from WIDER
FACE [24] to collect positives, negatives and part face. Then,
we crop faces from CelebA [23] as landmark faces
2) R-Net: We use first stage of our framework to detect faces
4
Fig. 3. (a) Validation loss of O-Net with and without hard sample mining. (b)
“JA” denotes joint face alignment learning while “No JA” denotes do not joint
it. “No JA in BBR” denotes do not joint it while training the CNN for bounding
box regression.
from WIDER FACE [24] to collect positives, negatives and
part face while landmark faces are detected from CelebA [23].
3) O-Net: Similar to R-Net to collect data but we use first two
stages of our framework to detect faces.
B. The effectiveness of online hard sample mining
To evaluate the contribution of the proposed online hard
sample mining strategy, we train two O-Nets (with and without
online hard sample mining) and compare their loss curves. To
make the comparison more directly, we only train the O-Nets
for the face classification task. All training parameters includ-
ing the network initialization are the same in these two O-Nets.
To compare them easier, we use fix learning rate. Fig. 3 (a)
shows the loss curves from two different training ways. It is
very clear that the hard sample mining is beneficial to perfor-
mance improvement.
C. The effectiveness of joint detection and alignment
To evaluate the contribution of joint detection and alignment,
we evaluate the performances of two different O-Nets (joint
facial landmarks regression task and do not joint it) on FDDB
(with the same P-Net and R-Net for fair comparison). We also
compare the performance of bounding box regression in these
two O-Nets. Fig. 3 (b) suggests that joint landmarks localiza-
tion task learning is beneficial for both face classification and
bounding box regression tasks.
D. Evaluation on face detection
To evaluate the performance of our face detection method,
we compare our method against the state-of-the-art methods [1,
5, 6, 11, 18, 19, 26, 27, 28, 29] in FDDB, and the
state-of-the-art methods [1, 24, 11] in WIDER FACE. Fig. 4
(a)-(d) shows that our method consistently outperforms all the
previous approaches by a large margin in both the benchmarks.
We also evaluate our approach on some challenge photos1.
E. Evaluation on face alignment
In this part, we compare the face alignment performance of
our method against the following methods: RCPR [12], TSPM
[7], Luxand face SDK [17], ESR [13], CDM [15], SDM [21],
and TCDCN [22]. In the testing phase, there are 13 images that
our method fails to detect face. So we crop the central region of
these 13 images and treat them as the input for O-Net. The
mean error is measured by the distances between the estimated
1 Examples are showed in http://kpzhang93.github.io/SPL/index.html
Fig. 4. (a) Evaluation on FDDB. (b-d) Evaluation on three subsets of WIDER
FACE. The number following the method indicates the average accuracy. (e)
Evaluation on AFLW for face alignment
landmarks and the ground truths, and normalized with respect
to the inter-ocular distance. Fig. 4 (e) shows that our method
outperforms all the state-of-the-art methods with a margin.
F. Runtime efficiency
Given the cascade structure, our method can achieve very fast
speed in joint face detection and alignment. It takes 16fps on a
2.60GHz CPU and 99fps on GPU (Nvidia Titan Black). Our
implementation is currently based on un-optimized MATLAB
code.
IV. CONCLUSION
In this paper, we have proposed a multi-task cascaded CNNs
based framework for joint face detection and alignment. Ex-
perimental results demonstrate that our methods consistently
outperform the state-of-the-art methods across several chal-
lenging benchmarks (including FDDB and WIDER FACE
benchmarks for face detection, and AFLW benchmark for face
alignment) while keeping real time performance. In the future,
we will exploit the inherent correlation between face detection
and other face analysis tasks, to further improve the perfor-
mance.
Advertisement
REFERENCES
[1] B. Yang, J. Yan, Z. Lei, and S. Z. Li, “Aggregate channel eatures for
multi-view face detection,” in IEEE International Joint Conference on
Biometrics, 2014, pp. 1-8.
5
part model for face detection,” in IEEE International Conference on Bio-
metrics Theory, Applications and Systems, 2015, pp. 1-8.
[28] G. Ghiasi, and C. C. Fowlkes, “Occlusion Coherence: Detecting and
Localizing Occluded Faces,” arXiv preprint arXiv:1506.08347.
[29] S. S. Farfade, M. J. Saberian, and L. J. Li, “Multi-view face detection using
deep convolutional neural networks,” in ACM on International Conference
on Multimedia Retrieval, 2015, pp. 643-650.
[2] P. Viola and M. J. Jones, “Robust real-time face detection. International
journal of computer vision,” vol. 57, no. 2, pp. 137-154, 2004
[3] M. T. Pham, Y. Gao, V. D. D. Hoang, and T. J. Cham, “Fast polygonal
integration and its application in extending haar-like features to improve
object detection,” in IEEE Conference on Computer Vision and Pattern
Recognition, 2010, pp. 942-949.
[4] Q. Zhu, M. C. Yeh, K. T. Cheng, and S. Avidan, “Fast human detection
using a cascade of histograms of oriented gradients,” in IEEE Computer
Conference on Computer Vision and Pattern Recognition, 2006, pp.
1491-1498.
[5] M. Mathias, R. Benenson, M. Pedersoli, and L. Van Gool, “Face detection
without bells and whistles,” in European Conference on Computer Vision,
2014, pp. 720-735.
J. Yan, Z. Lei, L. Wen, and S. Li, “The fastest deformable part model for
object detection,” in IEEE Conference on Computer Vision and Pattern
Recognition, 2014, pp. 2497-2504.
[6]
[7] X. Zhu, and D. Ramanan, “Face detection, pose estimation, and landmark
localization in the wild,” in IEEE Conference on Computer Vision and
Pattern Recognition, 2012, pp. 2879-2886.
[8] M. Köstinger, P. Wohlhart, P. M. Roth, and H. Bischof, “Annotated facial
landmarks in the wild: A large-scale, real-world database for facial land-
mark localization,” in IEEE Conference on Computer Vision and Pattern
Recognition Workshops, 2011, pp. 2144-2151.
[9] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification
with deep convolutional neural networks,” in Advances in neural infor-
mation processing systems, 2012, pp. 1097-1105.
[10] Y. Sun, Y. Chen, X. Wang, and X. Tang, “Deep learning face representa-
tion by joint identification-verification,” in Advances in Neural Infor-
mation Processing Systems, 2014, pp. 1988-1996.
[11] S. Yang, P. Luo, C. C. Loy, and X. Tang, “From facial parts responses to
face detection: A deep learning approach,” in IEEE International Confer-
ence on Computer Vision, 2015, pp. 3676-3684.
[12] X. P. Burgos-Artizzu, P. Perona, and P. Dollar, “Robust face landmark
estimation under occlusion,” in IEEE International Conference on Com-
puter Vision, 2013, pp. 1513-1520.
[13] X. Cao, Y. Wei, F. Wen, and J. Sun, “Face alignment by explicit shape
regression,” International Journal of Computer Vision, vol 107, no. 2, pp.
177-190, 2012.
[14] T. F. Cootes, G. J. Edwards, and C. J. Taylor, “Active appearance models,”
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23,
no. 6, pp. 681-685, 2001.
[15] X. Yu, J. Huang, S. Zhang, W. Yan, and D. Metaxas, “Pose-free facial
landmark fitting via optimized part mixtures and cascaded deformable
shape model,” in IEEE International Conference on Computer Vision,
2013, pp. 1944-1951.
[16] J. Zhang, S. Shan, M. Kan, and X. Chen, “Coarse-to-fine auto-encoder
networks (CFAN) for real-time face alignment,” in European Conference
on Computer Vision, 2014, pp. 1-16.
[17] Luxand Incorporated: Luxand face SDK, http://www.luxand.com/
[18] D. Chen, S. Ren, Y. Wei, X. Cao, and J. Sun, “Joint cascade face detection
and alignment,” in European Conference on Computer Vision, 2014, pp.
109-122.
[19] H. Li, Z. Lin, X. Shen, J. Brandt, and G. Hua, “A convolutional neural
network cascade for face detection,” in IEEE Conference on Computer
Vision and Pattern Recognition, 2015, pp. 5325-5334.
[20] C. Zhang, and Z. Zhang, “Improving multiview face detection with mul-
ti-task deep convolutional neural networks,” IEEE Winter Conference
on Applications of Computer Vision, 2014, pp. 1036-1041.
[21] X. Xiong, and F. Torre, “Supervised descent method and its applications to
face alignment,” in IEEE Conference on Computer Vision and Pattern
Recognition, 2013, pp. 532-539.
[22] Z. Zhang, P. Luo, C. C. Loy, and X. Tang, “Facial landmark detection by
deep multi-task learning,” in European Conference on Computer Vision,
2014, pp. 94-108.
[23] Z. Liu, P. Luo, X. Wang, and X. Tang, “Deep learning face attributes in the
wild,” in IEEE International Conference on Computer Vision, 2015, pp.
3730-3738.
[24] S. Yang, P. Luo, C. C. Loy, and X. Tang, “WIDER FACE: A Face Detec-
tion Benchmark”. arXiv preprint arXiv:1511.06523.
[25] V. Jain, and E. G. Learned-Miller, “FDDB: A benchmark for face detec-
tion in unconstrained settings,” Technical Report UMCS-2010-009, Uni-
versity of Massachusetts, Amherst, 2010.
[26] B. Yang, J. Yan, Z. Lei, and S. Z. Li, “Convolutional channel features,” in
IEEE International Conference on Computer Vision, 2015, pp. 82-90.
[27] R. Ranjan, V. M. Patel, and R. Chellappa, “A deep pyramid deformable