Face Model Compression by Distilling Knowledge from Neurons

• Author: Ping Luo , Zhenyao Zhu, Ziwei Liu , Xiaogang Wang , and Xiaoou Tang
• Origin: http://www.aaai.org/ocs/index.php/AAAI/AAAI16/paper/download/11977/12130
• Related:

Learning from Simulated and Unsupervised Images through Adversarial Training

• Author: Ashish Shrivastava, Tomas Pfister, Oncel Tuzel, Josh Susskind, Wenda Wang, Russ Webb Apple Inc.
• Origin: http://www.gitxiv.com/posts/LHjwXv5jNZFLgngC5/learning-from-simulated-and-unsupervised-images-through
• Related:
  • https://arxiv.org/abs/1612.07828
  • https://github.com/carpedm20/simulated-unsupervised-tensorflow

A Discriminative Feature Learning Approach for Deep Face Recognition

• Author: Yandong Wen, Kaipeng Zhang, Zhifeng Li and Yu Qiao
• Origin: http://ydwen.github.io/papers/WenECCV16.pdf
• Related:
  • https://github.com/ydwen/caffe-face
  • https://github.com/pangyupo/mxnet_center_loss

center loss

Pose-Aware Face Recognition in the Wild

• Author: Iacopo Masi, Stephen Rawls, G ́ erard Medioni, Prem Natarajan
• Origin: http://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Masi_Pose-Aware_Face_Recognition_CVPR_2016_paper.pdf
• Related:
  • 人脸识别 - Pose-Aware Face Recognition in the Wild

We propose a method to push the frontiers of unconstrained face recognition in the wild, focusing on the problem of extreme pose variations.

We leverage deep Convolutional Neural Networks (CNNs) to learn discrimi-native representations we call Pose-Aware Models (PAMs)using 500K images from the CASIA WebFace dataset.

Learning Deep Features via Congenerous Cosine Loss for Person Recognition

• Author: Yu Liu, Hongyang Li, Xiaogang Wang
• Origin: https://arxiv.org/abs/1702.06890
• Related:
  • https://github.com/sciencefans/coco_loss

DSD: Regularizing Deep Neural Networks with Dense-Sparse-Dense Training Flow

• Author: Song Han, Jeff Pool, Sharan Narang, Huizi Mao, Shijian Tang, Erich Elsen, Bryan Catanzaro, John Tran, William J. Dally
• Origin: https://arxiv.org/abs/1607.04381
• Related:
  • https://github.com/songhan/SqueezeNet-DSD-Training

Experiments show that DSD training can improve the performance of a wide range of CNN, RNN and LSTMs on the tasks of image classification, caption generation and speech recognition.

Efficient MRF Energy Propagation for Video Segmentation via Bilateral Filters

• Author: Ozan Sener, Kemal Ugur, A. Aydin Alatan
• Origin: https://arxiv.org/abs/1301.5356
• Related:
  • https://github.com/ozansener/VideoSegmentationBiExp

Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields

• Author: Zhe Cao, Tomas Simon, Shih-En Wei, Yaser Sheikh
• Origin: https://arxiv.org/abs/1611.08050
• Related:
  • GitXiv
  • demo video

Revisiting Batch Normalization For Practical Domain Adaptation

• Author: Yanghao Li, Naiyan Wang, Jianping Shi, Jiaying Liu, Xiaodi Hou
• Origin: https://arxiv.org/abs/1603.04779
• Related:

DeepPose: Human Pose Estimation via Deep Neural Networks

• Author: Alexander Toshev, Christian Szegedy
• Origin: http://www.gitxiv.com/posts/JpkGNjWzWQM2fF7Dn/deeppose-stg-1-on-tensorflow
• Related:
  • http://arxiv.org/abs/1312.4659
  • https://github.com/asanakoy/deeppose_tf

YodaNN: An Architecture for Ultra-Low Power Binary-Weight CNN Acceleration

• Author: Renzo Andri, Lukas Cavigelli, Davide Rossi, Luca Benini
• Origin: https://arxiv.org/abs/1606.05487
• Related:

Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks

• Author: Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, Yu Qiao
• Origin: https://kpzhang93.github.io/MTCNN_face_detection_alignment/index.html?from=groupmessage&isappinstalled=0
• Related:
  • kpzhang93/MTCNN_face_detection_alignment
  • https://github.com/foreverYoungGitHub/MTCNN : implemented with Caffe, C++ interface.
  • https://github.com/dlunion/mtcnn : 基于caffe的mtcnn训练实现
  • AlphaQi/MTCNN-light : the implementation of MTCNN with no framework
  • caffe version: https://github.com/CongWeilin/mtcnn-caffe
  • mxnet version: https://github.com/Seanlinx/mtcnn
  • https://github.com/blankWorld/MTCNN-Accelerate-Onet
  • https://github.com/rocking5566/face-detection-MTCNN-ncnn

Wasserstein GAN

• Author: Martin Arjovsky, Soumith Chintala, Léon Bottou
• Origin: http://www.gitxiv.com/posts/BLGXoByEcDgGF3rxx/wasserstein-gan
• Related:
  • https://arxiv.org/abs/1701.07875
  • https://github.com/martinarjovsky/WassersteinGAN

Dynamic Network Surgery for Efficient DNNs

• Author: Yiwen Guo, Anbang Yao, Yurong Chen
• Origin: https://arxiv.org/abs/1608.04493
• Related:
  • Learning both Weights and Connections for Efficient Neural Networks

we propose a novel network compression method called dynamic network surgery, which can remarkably reduce the network complexity by making on-the-fly connection pruning.

Distilling the Knowledge in a Neural Network

• Author: Geoffrey Hinton, Oriol Vinyals, Jeff Dean
• Origin: https://arxiv.org/abs/1503.02531
• Related:
  • https://github.com/wentianli/knowledge_distillation_caffe
  • 论文笔记 《Distilling the Knowledge in a Neural Network》

teacher network, student network

soft targets

Temperature

Large-Margin Softmax Loss for Convolutional Neural Networks

• Author: Weiyang Liu, Yandong Wen, Zhiding Yu, Meng Yang
• Origin: https://arxiv.org/abs/1612.02295
• Related:
  • https://github.com/wy1iu/LargeMargin_Softmax_Loss
  • https://github.com/luoyetx/mx-lsoftmax
  • 基于Caffe的Large Margin Softmax Loss的实现（上）
  • 基于Caffe的Large Margin Softmax Loss的实现（中）

Targeting Ultimate Accuracy: Face Recognition via Deep Embedding

• Author: Jingtuo Liu, Yafeng Deng, Tao Bai, Zhengping Wei, Chang Huang
• Origin: https://arxiv.org/abs/1506.07310
• Related:

Inverse Compositional Spatial Transformer Networks

• Author: Chen-Hsuan Lin, Simon Lucey
• Origin: https://arxiv.org/abs/1612.03897
• Related:
  • https://github.com/ericlin79119/IC-STN

We demonstrate that IC-STNs can achieve better performance than conventional STNs with less model capacity

Towards End-to-End Face Recognition through Alignment Learning

• Author: Yuanyi Zhong, Jiansheng Chen, Bo Huang
• Origin: https://arxiv.org/abs/1701.07174
• Related:

Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space

• Author: Anh Nguyen, Jason Yosinski, Yoshua Bengio, Alexey Dosovitskiy, Jeff Clune
• Origin: http://www.gitxiv.com/posts/Z8t29dEvMWYbfJjx2/plug-and-play-generative-networks-conditional-iterative
• Related:
  • https://arxiv.org/abs/1612.00005
  • https://github.com/Evolving-AI-Lab/ppgn

Pixel-Level Domain Transfer

• Author: Donggeun Yoo, Namil Kim, Sunggyun Park, Anthony S. Paek, In So Kweon
• Origin: http://www.gitxiv.com/posts/xQ5sX5aMpQHpkiBa5/pixel-level-domain-transfer
• Related:
  • LookBook dataset
  • Donggeun Yoo's blog

We present an image-conditional image generation model. The model transfers an input domain to a target domain in semantic level, and generates the target image in pixel level.

ShaResNet: reducing residual network parameter number by sharing weights

• Author: Alexandre Boulch
• Origin: https://arxiv.org/abs/1702.08782
• Related:
  • https://github.com/aboulch/sharesnet

Funnel-Structured Cascade for Multi-View Face Detection with Alignment-Awareness

• Author: Shuzhe Wu, Meina Kan, Zhenliang He, Shiguang Shan, Xilin Chen
• Origin: http://arxiv.org/abs/1609.07304v1
• Related:
  • seetaface/SeetaFaceEngine

we propose a novel funnel-structured cascade (FuSt) detection framework. In a coarse-to-fine flavor, our FuSt consists of, from top to bottom, 1) multiple view-specific fast LAB cascade for extremely quick face proposal, 2) multiple coarse MLP cascade for further candidate window verification, and 3) a unified fine MLP cascade with shape-indexed features for accurate face detection.

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

• Author: Christian Ledig, Lucas Theis, Ferenc Huszar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, Wenzhe Shi
• Origin: http://www.gitxiv.com/posts/5Z9gyGthtEhoroGzg/photo-realistic-single-image-super-resolution-using-a
• Related:
  • arXiv
  • https://github.com/leehomyc/Photo-Realistic-Super-Resoluton
  • tensorflow implementation: https://github.com/david-gpu/srez

Unrolled Generative Adversarial Networks

• Author: Luke Metz, Ben Poole, David Pfau, Jascha Sohl-Dickstein
• Origin: https://arxiv.org/abs/1611.02163
• Related:
  • https://github.com/poolio/unrolled_gan

Range Loss for Deep Face Recognition with Long-tail

• Author: Xiao Zhang, Zhiyuan Fang, Yandong Wen, Zhifeng Li, Yu Qiao
• Origin: https://arxiv.org/abs/1611.08976
• Related:
  • https://github.com/Charrin/RangeLoss-Caffe

Age and Gender Estimation of Unfiltered Faces

• Author: Eran Eidinger, Roee Enbar, Tal Hassner
• Origin: http://www.openu.ac.il/home/hassner/Adience/EidingerEnbarHassner_tifs.pdf
• Related:

TABLE I: Benchmarks for age and gender estimation from photos. With the exception of the FG-NET Aging and UIUCIFP-Y benchmarks, the table includes only benchmarks which are presently available online to the research community.

dataset

• FG-NET: (inclued in megaface FGNET2)
• MORPH [12]: https://ebill.uncw.edu/C20231_ustores/web/store_main.jsp?STOREID=4
• UIUC-IFP-Y Internal Aging [13], [14]: Not publicly available, yet may presumably be obtained from its authors.
• PubFig [3]: http://www.cs.columbia.edu/CAVE/databases/pubfig/download/
• Gallagher group photos[5] (Group Photos benchmark): http://chenlab.ece.cornell.edu/people/Andy/ImagesOfGroups.html
• VADANA [15]: https://www.eecis.udel.edu/wiki/vims/index.php/Main/VADANA (requestment needed)
• Adience: http://www.openu.ac.il/home/hassner/Adience/data.html#agegender

DeMeshNet: Blind Face Inpainting for Deep MeshFace Verification

• Author: Shu Zhang, Ran He, Tieniu Tan
• Origin: https://arxiv.org/abs/1611.05271
• Related:

Spatial Transformer Networks

• Author: Max Jaderberg, Karen Simonyan, Andrew Zisserman, Koray Kavukcuoglu
• Origin: http://www.gitxiv.com/posts/5WTXTLuEA4Hd8W84G/spatial-transformer-networks
• Related:
  • https://arxiv.org/abs/1506.02025
  • Implementation for Spatial Transformer Network for Caffe
  • Update + whole caffe
  • Zehaos/TFSTN(TF API 1.0) https://github.com/Zehaos/TFSTN
  • Tensorflow Implementation of Spatial Transformer Networks

Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer

• Author: Sergey Zagoruyko, Nikos Komodakis
• Origin: http://www.gitxiv.com/posts/3vyshBa8koJoYpYMy/improving-the-performance-of-convolutional-neural-networks
• Related:
  • https://arxiv.org/abs/1612.03928
  • https://github.com/szagoruyko/attention-transfer

We show that activation-based attention transfer gives better improvements than full activation transfer, and can be combined with knowledge distillation

Ordinal Regression with Multiple Output CNN for Age Estimation

• Author: Zhenxing Niu, Mo Zhou, Le Wang, Xinbo Gao, Gang Hua
• Origin: http://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Niu_Ordinal_Regression_With_CVPR_2016_paper.pdf
• Related:
  • https://github.com/luoyetx/OrdinalRegression
  • kongsicong: 对参考论文Ordinal Regression with Multiple Output CNN for Age Estimation理解
  • https://github.com/kongsicong/Age_recognition_OR
  • AFAD Dataset

we propose an End-to-End learning approach to address ordinal regression problems using deep Convolutional Neural Network, which could simultaneously conduct feature learning and regression modeling

we publish an Asian Face Age Dataset (AFAD) containing more than 160K facial images with precise age ground-truths, which is the largest public age dataset to date.

this is the first work to address ordinal regression problems by using CNN, and achieves the state-of-the-art performance on both the MORPH and AFAD datasets

dataset

• MORPH II: https://ebill.uncw.edu/C20231_ustores/web/store_main.jsp?STOREID=4
• AFAD:
  • http://afad-dataset.github.io/
  • https://github.com/afad-dataset/tarball

training, testing set

Following the experimental setting in [6][7][31], for both MORPH II and AFAD datasets, we randomly divide the whole dataset into two parts: one part (i.e., 80% of the whole data) is used for training, and the other one (i.e., 20% of the whole data) is used for testing.

performance

The performance is measured by the Mean Absolute Error (MAE) metric and the Cumulative Score (CS).

Learning from Noisy Labels with Distillation

• Author: Yuncheng Li, Jianchao Yang, Yale Song, Liangliang Cao, Jiebo Luo, Jia Li
• Origin: https://arxiv.org/abs/1703.02391v1
• Related:

we distill the knowledge learned the small clean dataset to facilitate learning a better model from the entire noisy dataset.

Continuous Supervised Descent Method for Facial Landmark Localisation

• Author: moliusimon
• Origin: unknown
• Related:
  • https://github.com/moliusimon/csdm

Age and Gender Classification using Convolutional Neural Networks

• Author: Gil Levi, Tal Hassner
• Origin: http://www.openu.ac.il/home/hassner/projects/cnn_agegender/
• Related:
  • paper: Age and Gender Classification using Convolutional Neural Networks
  • tensorflow: https://github.com/dpressel/rude-carnie
  • caffe: https://github.com/GilLevi/AgeGenderDeepLearning
  • caffe: https://github.com/eveningglow/age-and-gender-classification
  • cnn_age_gender_models_and_data.0.0.1.zip

we propose a simple convolutional net architecture that can be used even when the amount of learning data is limited.

We evaluate our method on the recent Adience benchmark for age and gender estimation and show it to dramatically outperform current state-of-the-art methods.

dataset (age)

• UIUC-IFP-Y [12, 19]: Not publicly available, yet may presumably be obtained from its authors.
• FG-NET [30]: (inclued in megaface FGNET2)
• MORPH [43]: https://ebill.uncw.edu/C20231_ustores/web/store_main.jsp?STOREID=4
• Group Photos benchmark [14]: http://chenlab.ece.cornell.edu/people/Andy/ImagesOfGroups.html
• Adience benchmark: http://www.openu.ac.il/home/hassner/Adience/data.html#agegender

dataset (gender)

• FERET benchmark [39]
• Adience benchmark

Residual Networks of Residual Networks: Multilevel Residual Networks

• Author: Ke Zhang, Miao Sun, Tony X. Han, Xingfang Yuan, Liru Guo, Tao Liu
• Origin: https://arxiv.org/abs/1608.02908v2
• Related:

Neural Aggregation Network for Video Face Recognition

• Author: Jiaolong Yang, Peiran Ren, Dong Chen, Fang Wen, Hongdong Li, Gang Hua
• Origin: https://arxiv.org/abs/1603.05474
• Related:

Face Spoof Detection with Image Distortion Analysis

• Author: D. Wen, H. Han, and A. K. Jain
• Origin: http://biometrics.cse.msu.edu/projects/face_recog_spoofing.html
• Related:
  • MSU Mobile Face Spoofing Database (MSU MFSD)

The public available MSU MFSD Database for face spoof attack consists of 280 video clips of photo and video attack attempts to 35 clients.

Convolutional Pose Machines

• Author: Shih-En Wei, Varun Ramakrishna, Takeo Kanade, Yaser Sheikh
• Origin: https://arxiv.org/abs/1602.00134
• Related:
  • https://github.com/shihenw/convolutional-pose-machines-release

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

• Author: Jiajun Wu, Chengkai Zhang, Tianfan Xue, William T. Freeman, Joshua B. Tenenbaum
• Origin: http://www.gitxiv.com/posts/5du5BTLcrgJBhgJgd/learning-a-probabilistic-latent-space-of-object-shapes-via
• Related:
  • http://3dgan.csail.mit.edu/

Unsupervised Cross-Domain Image Generation

• Author: Yaniv Taigman, Adam Polyak & Lior Wolf
• Origin: https://arxiv.org/pdf/1611.02200.pdf
• Related:
  • https://github.com/yunjey/dtn-tensorflow

We apply our method to visual domains including digits and face images and demon-strate its ability to generate convincing novel images of previously unseen entities, while preserving their identity.

Motion-based countermeasures to photo attacks in face recognition

• Author: Andre Anjos, Murali Mohan Chakka and S ´ ebastien Marcel
• Origin: http://publications.idiap.ch/downloads/papers/2013/Anjos_IETBIOMETRICS_2013.pdf
• Related:
  • 活体检测Face Anti-spoofing综述
  • Replay-Attack dataset
  • 人脸识别中的活体检测

The Replay-Attack Database for face spoofing consists of 1300 video clips of photo and video attack attempts to 50 clients, under different lighting conditions. This Database was produced at the Idiap Research Institute, in Switzerland.

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

• Author: Alec Radford, Luke Metz, Soumith Chintala
• Origin: http://www.gitxiv.com/posts/pNosr4Zrgn8uPEi9R/unsupervised-representation-learning-with-deep-convolutional
• Related:
  • 深度卷积对抗生成网络(DCGAN)
  • 生成式对抗网络GAN研究进展（五）——Deep Convolutional Generative Adversarial Nerworks，DCGAN
  • 生成式对抗网络GAN研究进展（二）——原始GAN
  • 对抗生成网络（Generative Adversarial Net)

Finding Tiny Faces

• Author: Peiyun Hu, Deva Ramanan
• Origin: https://www.cs.cmu.edu/~peiyunh/tiny/
• Related:
  • https://github.com/cydonia999/Tiny_Faces_in_Tensorflow
  • paper
  • https://github.com/peiyunh/tiny
  • A MXNet tiny face detector
  • data: human_data.zip
  • model: hr_res101.mat
  • WIDER FACE: A Face Detection Benchmark
  • FDDB: A Benchmark for Face Detection in Unconstrained Settings.

3D Human Pose Estimation = 2D Pose Estimation + Matching

• Author: Ching-Hang Chen, Deva Ramanan
• Origin: https://arxiv.org/abs/1612.06524
• Related:
  • https://github.com/flyawaychase/3DHumanPose

Visual Saliency Prediction with Generative Adversarial Networks

• Author: Junting Pan, Cristian Canton Ferrer, Kevin McGuinness, Noel E. O'Connor, Jordi Torres, Elisa Sayrol, Xavier Giro-i-Nieto
• Origin: http://www.gitxiv.com/posts/PriiZWZoPWDePEevp/salgan-visual-saliency-prediction-with-generative
• Related:
  • https://arxiv.org/abs/1701.01081
  • https://github.com/imatge-upc/saliency-salgan-2017

End-to-end people detection in crowded scenes

• Author: Russell Stewart, Mykhaylo Andriluka
• Origin: https://arxiv.org/abs/1506.04878
• Related:
  • https://github.com/Russell91/ReInspect

Why does deep and cheap learning work so well?

• Author: Henry W. Lin, Max Tegmark
• Origin: http://arxiv.org/abs/1608.08225
• Related:
  • fb message

we explore how these properties translate into exceptionally simple neural networks approximat-
ing both natural phenomena such as images and abstract representations thereof such as drawings.

We further argue that when the statistical process generating the data is of a certain hierarchical form prevalent in physics and machine-learning, a deep neural network can be more efficient than a shallow one.

Generative Adversarial Nets

• Author: Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio
• Origin: https://arxiv.org/abs/1406.2661
• Related:

Deep Identity-aware Transfer of Facial Attributes

• Author: Mu Li, Wangmeng Zuo, David Zhang
• Origin: https://arxiv.org/abs/1610.05586
• Related:

Binarized Convolutional Landmark Localizers for Human Pose Estimation and Face Alignment with Limited Resources

• Author: Adrian Bulat, Georgios Tzimiropoulos
• Origin: https://www.adrianbulat.com/binary-cnn-landmarks
• Related:
  • https://arxiv.org/abs/1703.00862
  • https://github.com/1adrianb/binary-human-pose-estimation