Efficient Deep Model Selection

Size: px
Start display at page:

Download "Efficient Deep Model Selection"

Transcription

1 Efficient Deep Model Selection Jose Alvarez Researcher Data61, CSIRO, Australia GTC, May 9 th

2

3 conv1 conv2 conv3 conv4 conv5 conv6 conv7 conv8 softmax prediction???????? Num Classes

4 Today s talk 1. It is possible to train more-efficient architectures without compromising accuracy 2. We can jointly learn the architecture and the parameters Additional benefits at train time.

5 Efficient Networks Convolutional Neural Networks (ConvNets)

6 150 Efficient Networks Num. Parameters (in Millions) LeNet AlexNet 2014 VGGNet-16

7 150 Efficient Networks Num. Parameters (in Millions) LeNet AlexNet 2014 VGGNet-16

8 150 Efficient Networks Num. Parameters (in Millions) LeNet AlexNet 2014 VGGNet-16

9 150 Efficient Networks Num. Parameters (in Millions) LeNet AlexNet 2014 VGGNet Residual-Nets More recent architectures? Residual Networks: Require 2 / 4 Titan-X (12GB) full capacity for training (memory requirements) 2 weeks training time (ResNet layers)

10 Efficient Networks TitanX GPU

11 Efficient Networks Embedded Platforms?

12 Efficient Networks Embedded Platforms?

13 Embedded Platforms Embedded devices with limited resources / power 2014 Jetson TK Jetson TX1

14 Efficient Networks TRAINING TESTING

15 Efficient Networks

16 Efficient Networks Jetson TK Limited resources / 2016 power / time TRAINING 2016 Jetson TX1 TESTING

17 Efficient Networks 2013 Larger data 2015 Spatio-temporal data video Hyper spectral Image Remote sensing 2016 TRAINING

18 Efficient Networks 2013 Larger data 2015 Spatio-temporal data video Hyper spectral Image Remote sensing 2016 Rapid prototyping TRAINING

19 Efficient Networks conv1 conv2 conv3 conv4 conv5 conv6 conv7 conv8 softmax prediction Num Classes TRAINING

20 Efficient Networks conv1 conv2 conv3 conv4 conv5 conv6 conv7 conv8 softmax prediction Num Classes TRAINING Additional TEST benefits

21 Talk Road Map Related work Efficient Networks DecomposeMe Model Selection Next Steps

22 Compacting ConvNets (related work)

23 Compacting ConvNets Test time Network distillation Network pruning Low rank approximations Train time Learning constrained filters Inception

24 Compacting ConvNets: At test time Network distillation Network pruning Low rank approximations

25 Compacting ConvNets: At test time Network distillation Large network learns from data Generate labels using the trained network Train smaller nets using the output or soft layer Geoffrey Hinton, Oriol Vinyals, Jeff Dean. Distilling the Knowledge in a Neural Network. NIPSw 2015

26 Compacting ConvNets: At test time Network distillation Network pruning Directly remove unimportant parameters during training (usually) Requires second derivatives. Remove parameters + quantification 1 Good compression rates (orthogonal to other approaches). S. Han, H. Mao, and W. J. Dally. Deep compression: Compressing deep neural network with pruning, trained quantization and huffman coding. ICLR 2016

27 Compacting ConvNets: At test time Network distillation Network pruning Low rank approximations

28 Compacting ConvNets: At test time Low rank approximations Weights are approximated by a combination of rank 1 tensors. Max Jaderberg, Andrea Vedaldi, Andrew Zisserman Speeding up Convolutional Neural Networks with Low Rank Expansions. BMVC 2014

29 Compacting ConvNets: At test time Weak-Points Needs a full-rank network completely trained Not all filters can be approximated Theoretical speeds-up with drop of performance. Emily Denton, Wojciech Zaremba, Joan Bruna, Yann LeCun, Rob Fergus. Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation. NIPS 2014

30 Compacting ConvNets: At train time

31 Compacting ConvNets: At train time Learning constrained filters Same receptive field but less parameters 49C 2 vs. 3x(3x3)C 2 K. Simonyan, A. Zisserman. Very Deep Convolutional Networks for Large-Scale Image Recognition. ICLR, 2015

32 Compacting ConvNets: At train time Learning constrained filters Same receptive field but less parameters Deeper networks (more non-linearities) K. Simonyan, A. Zisserman. Very Deep Convolutional Networks for Large-Scale Image Recognition. ICLR, 2015

33 Compacting ConvNets: At train time Inception modules Fewer convolutions and an expansion layer Szegedy, Going Deeper with Convolutions, CVPR2014

34 Compacting ConvNets: At train time Inception v3 Szegedy et al., Rethinking the inception Architecture for Computer Vision. CVPR2016

35 Model Selection

36 Model Selection Common Approach: empirical Set-up Empirically set the number of neurons. Prune neurons as a post-processing.

37 Model Selection Common Approach: empirical Set-up Learning-based approaches (difficult to scale up) Optimal Brain Damage. LeCun et al. NIPS 1991 Learning Structured Sparsity in Deep Neural Networks. Wen, Wu, Wang, Chen, and Li. NIPS 2016 Convolutional Neural Fabrics. Saxena and Verbeek. NIPS 2016

38 DecomposeMe Filter Compositions for End-to-End Learning

39 Filter Compositions for End-to-End Learning F= v 1 h 1T +v 2 h 2T + + v k h k T Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

40 Filter Compositions for End-to-End Learning Convolution layer (F filters) L ~ F d h

41 Filter Compositions for End-to-End Learning L < F

42 Filter Compositions for End-to-End Learning Key properties: Filter restrictions during training (low-rank). Larger receptive fields. Deeper models (ReLU): increased capacity. Additional parameter sharing. Reduced within filter parameter redundancy. Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

43 Filter Compositions for End-to-End Learning What have we learned? DecomposeMe (without non-linearity) AlexNet Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

44 Filter Compositions for End-to-End Learning What have we learned? DecomposeMe (without non-linearity) DecomposeMe Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

45 Classification Results

46 The Architecture Dec1 Dec2 Dec3 Dec4 Dec5 Dec6 Dec7 Dec8 FC 1000

47 Quantitative Results: ImageNet ImageNet dataset: 1.2 million training images and for validation split in 1000 categories. Between 5000 and training images per class. No data augmentation (random flip).

48 Quantitative Results: ImageNet ImageNet dataset: 1.2 million training images and for validation split in 1000 categories. Between 5000 and training images per class No data augmentation (random flip). NETWORK NUMBER OF PARAMETERS NUMBER OF CONV. LAYERS TOP-1 ACCURACY (CENTER CROP) AlexNet OWT Bn 61M % B-NET (VGG-B) 133M % OURS* 7.1M % Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

49 Computational Cost

50 Computational Cost Number of parameters Input Channels Output Channels Intermediate channels Kernel dimension Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

51 Time Computational Cost (time) 400 Forward (Inference) Time on Two TitanX B-Net (VGG-B) AlexNet OWT Bn Ours Batch Size Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

52 Time Computational Cost (time) Forward-backward Time on Two TitanX B-Net (VGG-B) AlexNet OWT Bn Ours Rapid prototyping ~10 hours using 4-Tesla M Batch Size Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

53 Residual Networks

54 Quantitative Results: ImageNet Residual Net decomposed -- model: 256-d 1x1, 64 3x1, 64 1x3, 64 1x1, 256 relu relu relu

55 Quantitative Results: ImageNet ImageNet dataset: 256-d 1x1, 64 3x1, 64 1x3, 64 1x1, 256 relu relu relu NETWORK TOP-1 ACCURACY (CENTER CROP) TOP-5 ACCURACY (CENTER CROP) ResNet % 93.3% ResNet-152-DEC 77.7% 93.7% Alvarez and Petersson, DecomposeMe: Simplifying ConvNets for End-to-End Learning. Arxiv 2016

56 Computational Cost (time) ResNet-101 (Relative improvement): 256-d 1x1, 64 relu 3x1, 64 relu 1x3, 64 relu 1x1, 256 NETWORK PARAMS FWD TIME (BATCH 8) TOP-5 ACC ResNet-101_ % 8.45% -0.7% ResNet-101_ % 34.4% -1.0% ResNet-101_ % 40.56% -1.5%

57 Generalization to other applications

58 Semantic Segmentation Building Tree Vehicle Side-walk Road Romera, Alvarez et al., Efficient ConvNet for Real-Time Semantic Segmentation. To appear in IEEE-IV 2017,

59 Semantic Segmentation Romera, Alvarez et al., Efficient ConvNet for Real-Time Semantic Segmentation. To appear in IEEE-IV 2017,

60 Model Selection prediction

61 Learning the Number of Neurons Our Approach: Pruning-aware training

62 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Directly reduce (select the optimum) the number of neurons. Significant memory reductions with performance improvements Still start from an over-parameterized network to help training. 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

63 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Weight Decay (prevent weights with large values) 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

64 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Neuron Convolutional kernel Convolutional layer 5x1x3x3 Considers each parameter independently Weight Decay (prevent weights with large values) 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

65 Learning the Number of Neurons Our Approach: Pruning-aware training 2 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

66 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Removed To be kept Size of the group 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

67 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Removed To be kept Direct benefits at test time (the complete kernel is removed) Size of the group 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

68 Learning the Number of Neurons Our Approach: Pruning-aware training 2 Removed To be kept 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

69 Learning the Number of Neurons Our Approach: Pruning-aware training 2 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

70 Learning the Number of Neurons Training Process: Proximal Operator + SGD Take a step with respect to the normal loss and then apply the proximal operator of the regularizer 3. Incremental learning (SGD) over the dataset (1 EPOCH) 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS Simon, Friedman, Hastie, Tibshirani. A sparse-group lasso. Journal of Computational and Graphical Statistics, 2013

71 Learning the Number of Neurons Training Process: Proximal Operator + SGD Take a step with respect to the normal loss and then apply the proximal operator of the regularizer 3. Proximal operator for SGS 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS Simon, Friedman, Hastie, Tibshirani. A sparse-group lasso. Journal of Computational and Graphical Statistics, 2013

72 Learning the Number of Neurons Our Approach: Pruning-aware training Projection effect during training

73 Classification Results

74 Learning the Number of Neurons Quantitative Results on ImageNet Train an over-complete architecture up to 768 neurons per layer (Dec 8-768) Dec Dec Dec Dec Dec Dec Dec Dec FC Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

75 Learning the Number of Neurons Quantitative Results on ImageNet 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

76 Applicable to New DataSets?

77 Learning the Number of Neurons ICDAR2003: Character Recognition in camera captured images TESCO, Value Washing Up Liquid PEPSI 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016 The Rab Butler Building

78 Learning the Number of Neurons Quantitative Results on ICDAR2003 Train an over-complete architecture up to 512 neurons per layer (Dec 3 ) Over Parameterization Dec Dec Dec FC 36 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

79 Learning the Number of Neurons Results on ICDAR2003 Character Recognition Dataset 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

80 Computational Cost (test time)

81 Learning the Number of Neurons Additional benefits at test time 2 : speeds up and memory savings 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

82 Learning the Number of Neurons Additional benefits at test time 2 : Feature Extraction In the last layer 2 Alvarez and Salzmann, Learning the Number of Neurons in Deep Networks, NIPS 2016

83 Number of Layers

84 Learning the Layers Skip connection Dec1 Dec2 Dec3 Dec4 Dec5 Dec6 Dec7 Dec7-1 Dec7-2 Dec8 Dec8-1 Dec8-2 FC Skip connection 1000

85 Number of neurons Learning the Layers Skip connection Dec1 Dec2 Dec3 Dec4 Dec5 Dec6 Dec7 Dec7-1 Dec7-2 Dec8 Dec8-1 Dec8-2 FC Skip connection Initial number Learned number L1v L1h L2v L2h L3v L3h L4v L4h L5v L5h L6v L6h L7v L7hL7-1vL7-1hL7-2vL7-2hL8v L8hL8-1vL8-1hL8-2vL8-2h Layer Name

86 Number of neurons Learning the Layers Skip connection Dec1 Dec2 Dec3 Dec4 Dec5 Dec6 Dec7 Dec7-1 Dec7-2 Dec8 Dec8-1 Dec8-2 FC Skip connection Initial number Learned number No impact in performance L1v L1h L2v L2h L3v L3h L4v L4h L5v L5h L6v L6h L7v L7hL7-1vL7-1hL7-2vL7-2hL8v L8hL8-1vL8-1hL8-2vL8-2h Layer Name

87 Training Efficient

88 Improving Training Efficiency Projection effect during training Reload the model Change in the LR 70% train speed-up (ICDAR dataset)

89 Summary

90 Summary It is possible to train more-efficient architectures without compromising accuracy based on 1-D convolution kernels:

91 Summary It is possible to train more-efficient architectures without compromising accuracy based on 1-D convolution kernels. We can jointly learn the architecture and the parameters using structure-sparsity regularization: conv1 conv2 conv3 conv4 conv5 conv6 conv7 conv8 softmax prediction Num Classes TRAINING Additional TEST benefits

92 Summary It is possible to train more-efficient architectures without compromising accuracy based on 1-D convolution kernels. We can jointly learn the architecture and the parameters using structure-sparsity regularization. Additional benefits at training time.

93 Next Steps

94 Next Steps Increase performance unfreezing neurons: Naive unfreezing all neurons at the end seems to be a Bad idea.

95 Next Steps Increase performance unfreezing neurons. Add post-processing steps to reduce computational costs: Additional L1-pruning: Additional pruning of the learned model shows further benefits.

96 Next Steps Increase performance unfreezing neurons Add post-processing steps to reduce computational costs: Weight quantization: We are currently able to reduce to 3 bits with minor lost in performance.

97 Next Steps Increase performance unfreezing neurons Add post-processing steps to reduce computational costs. Design tree-structures to learn more complex layer architectures.

98 Thank you Jose Alvarez Researcher Data61, CSIRO, Australia

CSE Introduction to High-Perfomance Deep Learning ImageNet & VGG. Jihyung Kil

CSE Introduction to High-Perfomance Deep Learning ImageNet & VGG. Jihyung Kil CSE 5194.01 - Introduction to High-Perfomance Deep Learning ImageNet & VGG Jihyung Kil ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton,

More information

Deep Compression and EIE: Efficient Inference Engine on Compressed Deep Neural Network

Deep Compression and EIE: Efficient Inference Engine on Compressed Deep Neural Network Deep Compression and EIE: Efficient Inference Engine on Deep Neural Network Song Han*, Xingyu Liu, Huizi Mao, Jing Pu, Ardavan Pedram, Mark Horowitz, Bill Dally Stanford University Our Prior Work: Deep

More information

Automated diagnosis of pneumothorax using an ensemble of convolutional neural networks with multi-sized chest radiography images

Automated diagnosis of pneumothorax using an ensemble of convolutional neural networks with multi-sized chest radiography images Automated diagnosis of pneumothorax using an ensemble of convolutional neural networks with multi-sized chest radiography images Tae Joon Jun, Dohyeun Kim, and Daeyoung Kim School of Computing, KAIST,

More information

Multi-attention Guided Activation Propagation in CNNs

Multi-attention Guided Activation Propagation in CNNs Multi-attention Guided Activation Propagation in CNNs Xiangteng He and Yuxin Peng (B) Institute of Computer Science and Technology, Peking University, Beijing, China pengyuxin@pku.edu.cn Abstract. CNNs

More information

Convolutional Neural Networks (CNN)

Convolutional Neural Networks (CNN) Convolutional Neural Networks (CNN) Algorithm and Some Applications in Computer Vision Luo Hengliang Institute of Automation June 10, 2014 Luo Hengliang (Institute of Automation) Convolutional Neural Networks

More information

Hierarchical Convolutional Features for Visual Tracking

Hierarchical Convolutional Features for Visual Tracking Hierarchical Convolutional Features for Visual Tracking Chao Ma Jia-Bin Huang Xiaokang Yang Ming-Husan Yang SJTU UIUC SJTU UC Merced ICCV 2015 Background Given the initial state (position and scale), estimate

More information

arxiv: v1 [stat.ml] 23 Jan 2017

arxiv: v1 [stat.ml] 23 Jan 2017 Learning what to look in chest X-rays with a recurrent visual attention model arxiv:1701.06452v1 [stat.ml] 23 Jan 2017 Petros-Pavlos Ypsilantis Department of Biomedical Engineering King s College London

More information

CS-E Deep Learning Session 4: Convolutional Networks

CS-E Deep Learning Session 4: Convolutional Networks CS-E4050 - Deep Learning Session 4: Convolutional Networks Jyri Kivinen Aalto University 23 September 2015 Credits: Thanks to Tapani Raiko for slides material. CS-E4050 - Deep Learning Session 4: Convolutional

More information

Y-Net: Joint Segmentation and Classification for Diagnosis of Breast Biopsy Images

Y-Net: Joint Segmentation and Classification for Diagnosis of Breast Biopsy Images Y-Net: Joint Segmentation and Classification for Diagnosis of Breast Biopsy Images Sachin Mehta 1, Ezgi Mercan 1, Jamen Bartlett 2, Donald Weaver 2, Joann G. Elmore 1, and Linda Shapiro 1 1 University

More information

arxiv: v2 [cs.cv] 22 Mar 2018

arxiv: v2 [cs.cv] 22 Mar 2018 Deep saliency: What is learnt by a deep network about saliency? Sen He 1 Nicolas Pugeault 1 arxiv:1801.04261v2 [cs.cv] 22 Mar 2018 Abstract Deep convolutional neural networks have achieved impressive performance

More information

An Artificial Neural Network Architecture Based on Context Transformations in Cortical Minicolumns

An Artificial Neural Network Architecture Based on Context Transformations in Cortical Minicolumns An Artificial Neural Network Architecture Based on Context Transformations in Cortical Minicolumns 1. Introduction Vasily Morzhakov, Alexey Redozubov morzhakovva@gmail.com, galdrd@gmail.com Abstract Cortical

More information

Network Dissection: Quantifying Interpretability of Deep Visual Representation

Network Dissection: Quantifying Interpretability of Deep Visual Representation Name: Pingchuan Ma Student number: 3526400 Date: August 19, 2018 Seminar: Explainable Machine Learning Lecturer: PD Dr. Ullrich Köthe SS 2018 Quantifying Interpretability of Deep Visual Representation

More information

Object Detectors Emerge in Deep Scene CNNs

Object Detectors Emerge in Deep Scene CNNs Object Detectors Emerge in Deep Scene CNNs Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, Antonio Torralba Presented By: Collin McCarthy Goal: Understand how objects are represented in CNNs Are

More information

Flexible, High Performance Convolutional Neural Networks for Image Classification

Flexible, High Performance Convolutional Neural Networks for Image Classification Flexible, High Performance Convolutional Neural Networks for Image Classification Dan C. Cireşan, Ueli Meier, Jonathan Masci, Luca M. Gambardella, Jürgen Schmidhuber IDSIA, USI and SUPSI Manno-Lugano,

More information

A convolutional neural network to classify American Sign Language fingerspelling from depth and colour images

A convolutional neural network to classify American Sign Language fingerspelling from depth and colour images A convolutional neural network to classify American Sign Language fingerspelling from depth and colour images Ameen, SA and Vadera, S http://dx.doi.org/10.1111/exsy.12197 Title Authors Type URL A convolutional

More information

Dual Path Network and Its Applications

Dual Path Network and Its Applications Learning and Vision Group (NUS), ILSVRC 2017 - CLS-LOC & DET tasks Dual Path Network and Its Applications National University of Singapore: Yunpeng Chen, Jianan Li, Huaxin Xiao, Jianshu Li, Xuecheng Nie,

More information

Rich feature hierarchies for accurate object detection and semantic segmentation

Rich feature hierarchies for accurate object detection and semantic segmentation Rich feature hierarchies for accurate object detection and semantic segmentation Ross Girshick, Jeff Donahue, Trevor Darrell, Jitendra Malik UC Berkeley Tech Report @ http://arxiv.org/abs/1311.2524! Detection

More information

B657: Final Project Report Holistically-Nested Edge Detection

B657: Final Project Report Holistically-Nested Edge Detection B657: Final roject Report Holistically-Nested Edge Detection Mingze Xu & Hanfei Mei May 4, 2016 Abstract Holistically-Nested Edge Detection (HED), which is a novel edge detection method based on fully

More information

HHS Public Access Author manuscript Med Image Comput Comput Assist Interv. Author manuscript; available in PMC 2018 January 04.

HHS Public Access Author manuscript Med Image Comput Comput Assist Interv. Author manuscript; available in PMC 2018 January 04. Discriminative Localization in CNNs for Weakly-Supervised Segmentation of Pulmonary Nodules Xinyang Feng 1, Jie Yang 1, Andrew F. Laine 1, and Elsa D. Angelini 1,2 1 Department of Biomedical Engineering,

More information

The Impact of Visual Saliency Prediction in Image Classification

The Impact of Visual Saliency Prediction in Image Classification Dublin City University Insight Centre for Data Analytics Universitat Politecnica de Catalunya Escola Tècnica Superior d Enginyeria de Telecomunicacions de Barcelona Eric Arazo Sánchez The Impact of Visual

More information

On Training of Deep Neural Network. Lornechen

On Training of Deep Neural Network. Lornechen On Training of Deep Neural Network Lornechen 2016.04.20 1 Outline Introduction Layer-wise Pre-training & Fine-tuning Activation Function Initialization Method Advanced Layers and Nets 2 Neural Network

More information

Convolutional Neural Networks for Text Classification

Convolutional Neural Networks for Text Classification Convolutional Neural Networks for Text Classification Sebastian Sierra MindLab Research Group July 1, 2016 ebastian Sierra (MindLab Research Group) NLP Summer Class July 1, 2016 1 / 32 Outline 1 What is

More information

Patch-based Head and Neck Cancer Subtype Classification

Patch-based Head and Neck Cancer Subtype Classification Patch-based Head and Neck Cancer Subtype Classification Wanyi Qian, Guoli Yin, Frances Liu, Advisor: Olivier Gevaert, Mu Zhou, Kevin Brennan Stanford University wqian2@stanford.edu, guoliy@stanford.edu,

More information

Automatic Detection of Knee Joints and Quantification of Knee Osteoarthritis Severity using Convolutional Neural Networks

Automatic Detection of Knee Joints and Quantification of Knee Osteoarthritis Severity using Convolutional Neural Networks Automatic Detection of Knee Joints and Quantification of Knee Osteoarthritis Severity using Convolutional Neural Networks Joseph Antony 1, Kevin McGuinness 1, Kieran Moran 1,2 and Noel E O Connor 1 Insight

More information

Exploratory Study on Direct Prediction of Diabetes using Deep Residual Networks

Exploratory Study on Direct Prediction of Diabetes using Deep Residual Networks Exploratory Study on Direct Prediction of Diabetes using Deep Residual Networks Samaneh Abbasi-Sureshjani, Behdad Dashtbozorg, Bart M. ter Haar Romeny, and François Fleuret Abstract Diabetes is threatening

More information

Retinopathy Net. Alberto Benavides Robert Dadashi Neel Vadoothker

Retinopathy Net. Alberto Benavides Robert Dadashi Neel Vadoothker Retinopathy Net Alberto Benavides Robert Dadashi Neel Vadoothker Motivation We were interested in applying deep learning techniques to the field of medical imaging Field holds a lot of promise and can

More information

Attentional Masking for Pre-trained Deep Networks

Attentional Masking for Pre-trained Deep Networks Attentional Masking for Pre-trained Deep Networks IROS 2017 Marcus Wallenberg and Per-Erik Forssén Computer Vision Laboratory Department of Electrical Engineering Linköping University 2014 2017 Per-Erik

More information

Elad Hoffer*, Itay Hubara*, Daniel Soudry

Elad Hoffer*, Itay Hubara*, Daniel Soudry Train longer, generalize better: closing the generalization gap in large batch training of neural networks Elad Hoffer*, Itay Hubara*, Daniel Soudry *Equal contribution Better models - parallelization

More information

Facial Expression Classification Using Convolutional Neural Network and Support Vector Machine

Facial Expression Classification Using Convolutional Neural Network and Support Vector Machine Facial Expression Classification Using Convolutional Neural Network and Support Vector Machine Valfredo Pilla Jr, André Zanellato, Cristian Bortolini, Humberto R. Gamba and Gustavo Benvenutti Borba Graduate

More information

arxiv: v2 [cs.cv] 31 Oct 2017

arxiv: v2 [cs.cv] 31 Oct 2017 Published as a conference paper at CVPR Workshop on Biometrics (2017) Deep LDA-Pruned Nets for Efficient Facial Gender Classification Qing Tian, Tal Arbel and James J. Clark Centre for Intelligent Machines

More information

Less is More: Culling the Training Set to Improve Robustness of Deep Neural Networks

Less is More: Culling the Training Set to Improve Robustness of Deep Neural Networks Less is More: Culling the Training Set to Improve Robustness of Deep Neural Networks Yongshuai Liu, Jiyu Chen, and Hao Chen University of California, Davis {yshliu, jiych, chen}@ucdavis.edu Abstract. Deep

More information

arxiv: v1 [cs.cv] 17 Aug 2017

arxiv: v1 [cs.cv] 17 Aug 2017 Deep Learning for Medical Image Analysis Mina Rezaei, Haojin Yang, Christoph Meinel Hasso Plattner Institute, Prof.Dr.Helmert-Strae 2-3, 14482 Potsdam, Germany {mina.rezaei,haojin.yang,christoph.meinel}@hpi.de

More information

Automatic Quality Assessment of Cardiac MRI

Automatic Quality Assessment of Cardiac MRI Automatic Quality Assessment of Cardiac MRI Ilkay Oksuz 02.05.2018 Contact: ilkay.oksuz@kcl.ac.uk http://kclmmag.org 1 Cardiac MRI Quality Issues Need for high quality images Wide range of artefacts Manual

More information

Convolutional Neural Networks for Estimating Left Ventricular Volume

Convolutional Neural Networks for Estimating Left Ventricular Volume Convolutional Neural Networks for Estimating Left Ventricular Volume Ryan Silva Stanford University rdsilva@stanford.edu Maksim Korolev Stanford University mkorolev@stanford.edu Abstract End-systolic and

More information

Differentiating Tumor and Edema in Brain Magnetic Resonance Images Using a Convolutional Neural Network

Differentiating Tumor and Edema in Brain Magnetic Resonance Images Using a Convolutional Neural Network Original Article Differentiating Tumor and Edema in Brain Magnetic Resonance Images Using a Convolutional Neural Network Aida Allahverdi 1, Siavash Akbarzadeh 1, Alireza Khorrami Moghaddam 2, Armin Allahverdy

More information

arxiv: v1 [cs.cv] 25 Jan 2018

arxiv: v1 [cs.cv] 25 Jan 2018 1 Convolutional Invasion and Expansion Networks for Tumor Growth Prediction Ling Zhang, Le Lu, Senior Member, IEEE, Ronald M. Summers, Electron Kebebew, and Jianhua Yao arxiv:1801.08468v1 [cs.cv] 25 Jan

More information

CS6501: Deep Learning for Visual Recognition. GenerativeAdversarial Networks (GANs)

CS6501: Deep Learning for Visual Recognition. GenerativeAdversarial Networks (GANs) CS6501: Deep Learning for Visual Recognition GenerativeAdversarial Networks (GANs) Today s Class Adversarial Examples Input Optimization Generative Adversarial Networks (GANs) Conditional GANs Style-Transfer

More information

arxiv: v2 [cs.cv] 19 Dec 2017

arxiv: v2 [cs.cv] 19 Dec 2017 An Ensemble of Deep Convolutional Neural Networks for Alzheimer s Disease Detection and Classification arxiv:1712.01675v2 [cs.cv] 19 Dec 2017 Jyoti Islam Department of Computer Science Georgia State University

More information

arxiv: v1 [cs.cr] 1 Nov 2018

arxiv: v1 [cs.cr] 1 Nov 2018 IMPROVING ADVERSARIAL ROBUSTNESS BY ENCOURAGING DISCRIMINATIVE FEATURES Chirag Agarwal Anh Nguyen Dan Schonfeld University of Illinois at Chicago, Chicago, IL, USA 60607 Auburn University, Auburn, AL,

More information

arxiv: v1 [cs.cv] 21 Nov 2018

arxiv: v1 [cs.cv] 21 Nov 2018 Rethinking ImageNet Pre-training Kaiming He Ross Girshick Piotr Dollár Facebook AI Research (FAIR) arxiv:1811.8883v1 [cs.cv] 21 Nov 218 Abstract We report competitive results on object detection and instance

More information

Classification of breast cancer histology images using transfer learning

Classification of breast cancer histology images using transfer learning Classification of breast cancer histology images using transfer learning Sulaiman Vesal 1 ( ), Nishant Ravikumar 1, AmirAbbas Davari 1, Stephan Ellmann 2, Andreas Maier 1 1 Pattern Recognition Lab, Friedrich-Alexander-Universität

More information

COMP9444 Neural Networks and Deep Learning 5. Convolutional Networks

COMP9444 Neural Networks and Deep Learning 5. Convolutional Networks COMP9444 Neural Networks and Deep Learning 5. Convolutional Networks Textbook, Sections 6.2.2, 6.3, 7.9, 7.11-7.13, 9.1-9.5 COMP9444 17s2 Convolutional Networks 1 Outline Geometry of Hidden Unit Activations

More information

arxiv: v2 [cs.cv] 29 Jan 2019

arxiv: v2 [cs.cv] 29 Jan 2019 Comparison of Deep Learning Approaches for Multi-Label Chest X-Ray Classification Ivo M. Baltruschat 1,2,, Hannes Nickisch 3, Michael Grass 3, Tobias Knopp 1,2, and Axel Saalbach 3 arxiv:1803.02315v2 [cs.cv]

More information

arxiv: v3 [cs.cv] 12 Feb 2017

arxiv: v3 [cs.cv] 12 Feb 2017 Published as a conference paper at ICLR 17 PAYING MORE ATTENTION TO ATTENTION: IMPROVING THE PERFORMANCE OF CONVOLUTIONAL NEURAL NETWORKS VIA ATTENTION TRANSFER Sergey Zagoruyko, Nikos Komodakis Université

More information

Highly Accurate Brain Stroke Diagnostic System and Generative Lesion Model. Junghwan Cho, Ph.D. CAIDE Systems, Inc. Deep Learning R&D Team

Highly Accurate Brain Stroke Diagnostic System and Generative Lesion Model. Junghwan Cho, Ph.D. CAIDE Systems, Inc. Deep Learning R&D Team Highly Accurate Brain Stroke Diagnostic System and Generative Lesion Model Junghwan Cho, Ph.D. CAIDE Systems, Inc. Deep Learning R&D Team Established in September, 2016 at 110 Canal st. Lowell, MA 01852,

More information

Motivation: Attention: Focusing on specific parts of the input. Inspired by neuroscience.

Motivation: Attention: Focusing on specific parts of the input. Inspired by neuroscience. Outline: Motivation. What s the attention mechanism? Soft attention vs. Hard attention. Attention in Machine translation. Attention in Image captioning. State-of-the-art. 1 Motivation: Attention: Focusing

More information

Skin cancer reorganization and classification with deep neural network

Skin cancer reorganization and classification with deep neural network Skin cancer reorganization and classification with deep neural network Hao Chang 1 1. Department of Genetics, Yale University School of Medicine 2. Email: changhao86@gmail.com Abstract As one kind of skin

More information

Neuromorphic convolutional recurrent neural network for road safety or safety near the road

Neuromorphic convolutional recurrent neural network for road safety or safety near the road Neuromorphic convolutional recurrent neural network for road safety or safety near the road WOO-SUP HAN 1, IL SONG HAN 2 1 ODIGA, London, U.K. 2 Korea Advanced Institute of Science and Technology, Daejeon,

More information

DeepASL: Enabling Ubiquitous and Non-Intrusive Word and Sentence-Level Sign Language Translation

DeepASL: Enabling Ubiquitous and Non-Intrusive Word and Sentence-Level Sign Language Translation DeepASL: Enabling Ubiquitous and Non-Intrusive Word and Sentence-Level Sign Language Translation Biyi Fang Michigan State University ACM SenSys 17 Nov 6 th, 2017 Biyi Fang (MSU) Jillian Co (MSU) Mi Zhang

More information

A CONVOLUTION NEURAL NETWORK ALGORITHM FOR BRAIN TUMOR IMAGE SEGMENTATION

A CONVOLUTION NEURAL NETWORK ALGORITHM FOR BRAIN TUMOR IMAGE SEGMENTATION A CONVOLUTION NEURAL NETWORK ALGORITHM FOR BRAIN TUMOR IMAGE SEGMENTATION 1 Priya K, 2 Dr.O. Saraniya 1 PG Scholar, 2 Assistant Professor Department Of ECE Government College of Technology, Coimbatore,

More information

Final Report: Automated Semantic Segmentation of Volumetric Cardiovascular Features and Disease Assessment

Final Report: Automated Semantic Segmentation of Volumetric Cardiovascular Features and Disease Assessment Final Report: Automated Semantic Segmentation of Volumetric Cardiovascular Features and Disease Assessment Tony Lindsey 1,3, Xiao Lu 1 and Mojtaba Tefagh 2 1 Department of Biomedical Informatics, Stanford

More information

Deep Learning Human Mind for Automated Visual Classification

Deep Learning Human Mind for Automated Visual Classification Deep Learning Human Mind for Automated Visual Classification C. Spampinato, S. Palazzo, I. Kavasidis, D. Giordano Department of Electrical, Electronics and Computer Engineering - PeRCeiVe Lab Viale Andrea

More information

Deep Learning Human Mind for Automated Visual Classification

Deep Learning Human Mind for Automated Visual Classification Deep Learning Human Mind for Automated Visual Classification C. Spampinato, S. Palazzo, I. Kavasidis, D. Giordano Department of Electrical, Electronics and Computer Engineering - PeRCeiVe Lab Viale Andrea

More information

Computational modeling of visual attention and saliency in the Smart Playroom

Computational modeling of visual attention and saliency in the Smart Playroom Computational modeling of visual attention and saliency in the Smart Playroom Andrew Jones Department of Computer Science, Brown University Abstract The two canonical modes of human visual attention bottomup

More information

Image-Based Estimation of Real Food Size for Accurate Food Calorie Estimation

Image-Based Estimation of Real Food Size for Accurate Food Calorie Estimation Image-Based Estimation of Real Food Size for Accurate Food Calorie Estimation Takumi Ege, Yoshikazu Ando, Ryosuke Tanno, Wataru Shimoda and Keiji Yanai Department of Informatics, The University of Electro-Communications,

More information

The Nottingham eprints service makes this work by researchers of the University of Nottingham available open access under the following conditions.

The Nottingham eprints service makes this work by researchers of the University of Nottingham available open access under the following conditions. Bulat, Adrian and Tzimiropoulos, Georgios (2016) Human pose estimation via convolutional part heatmap regression. In: 14th European Conference on Computer Vision (EECV 2016), 8-16 October 2016, Amsterdam,

More information

DEEP CONVOLUTIONAL ACTIVATION FEATURES FOR LARGE SCALE BRAIN TUMOR HISTOPATHOLOGY IMAGE CLASSIFICATION AND SEGMENTATION

DEEP CONVOLUTIONAL ACTIVATION FEATURES FOR LARGE SCALE BRAIN TUMOR HISTOPATHOLOGY IMAGE CLASSIFICATION AND SEGMENTATION DEEP CONVOLUTIONAL ACTIVATION FEATURES FOR LARGE SCALE BRAIN TUMOR HISTOPATHOLOGY IMAGE CLASSIFICATION AND SEGMENTATION Yan Xu1,2, Zhipeng Jia2,, Yuqing Ai2,, Fang Zhang2,, Maode Lai4, Eric I-Chao Chang2

More information

Quantifying Radiographic Knee Osteoarthritis Severity using Deep Convolutional Neural Networks

Quantifying Radiographic Knee Osteoarthritis Severity using Deep Convolutional Neural Networks Quantifying Radiographic Knee Osteoarthritis Severity using Deep Convolutional Neural Networks Joseph Antony, Kevin McGuinness, Noel E O Connor, Kieran Moran Insight Centre for Data Analytics, Dublin City

More information

Learning Convolutional Neural Networks for Graphs

Learning Convolutional Neural Networks for Graphs GA-65449 Learning Convolutional Neural Networks for Graphs Mathias Niepert Mohamed Ahmed Konstantin Kutzkov NEC Laboratories Europe Representation Learning for Graphs Telecom Safety Transportation Industry

More information

Weakly Supervised Coupled Networks for Visual Sentiment Analysis

Weakly Supervised Coupled Networks for Visual Sentiment Analysis Weakly Supervised Coupled Networks for Visual Sentiment Analysis Jufeng Yang, Dongyu She,Yu-KunLai,PaulL.Rosin, Ming-Hsuan Yang College of Computer and Control Engineering, Nankai University, Tianjin,

More information

NEONATAL SEIZURE DETECTION USING CONVOLUTIONAL NEURAL NETWORKS. Alison O Shea, Gordon Lightbody, Geraldine Boylan, Andriy Temko

NEONATAL SEIZURE DETECTION USING CONVOLUTIONAL NEURAL NETWORKS. Alison O Shea, Gordon Lightbody, Geraldine Boylan, Andriy Temko 2017 IEEE INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING, SEPT. 25 28, 2017, TOKYO, JAPAN NEONATAL SEIZURE DETECTION USING CONVOLUTIONAL NEURAL NETWORKS Alison O Shea, Gordon Lightbody,

More information

Context aware decision support in neurosurgical oncology based on an efficient classification of endomicroscopic data

Context aware decision support in neurosurgical oncology based on an efficient classification of endomicroscopic data International Journal of Computer Assisted Radiology and Surgery (2018) 13:1187 1199 https://doi.org/10.1007/s11548-018-1806-7 ORIGINAL ARTICLE Context aware decision support in neurosurgical oncology

More information

arxiv: v1 [cs.cv] 24 Jul 2018

arxiv: v1 [cs.cv] 24 Jul 2018 Multi-Class Lesion Diagnosis with Pixel-wise Classification Network Manu Goyal 1, Jiahua Ng 2, and Moi Hoon Yap 1 1 Visual Computing Lab, Manchester Metropolitan University, M1 5GD, UK 2 University of

More information

Shu Kong. Department of Computer Science, UC Irvine

Shu Kong. Department of Computer Science, UC Irvine Ubiquitous Fine-Grained Computer Vision Shu Kong Department of Computer Science, UC Irvine Outline 1. Problem definition 2. Instantiation 3. Challenge 4. Fine-grained classification with holistic representation

More information

Medical Image Analysis

Medical Image Analysis Medical Image Analysis 1 Co-trained convolutional neural networks for automated detection of prostate cancer in multiparametric MRI, 2017, Medical Image Analysis 2 Graph-based prostate extraction in t2-weighted

More information

Supplementary Online Content

Supplementary Online Content Supplementary Online Content Ting DS, Cheung CY-L, Lim G, et al. Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic

More information

Image Captioning using Reinforcement Learning. Presentation by: Samarth Gupta

Image Captioning using Reinforcement Learning. Presentation by: Samarth Gupta Image Captioning using Reinforcement Learning Presentation by: Samarth Gupta 1 Introduction Summary Supervised Models Image captioning as RL problem Actor Critic Architecture Policy Gradient architecture

More information

arxiv: v2 [cs.cv] 7 Jun 2018

arxiv: v2 [cs.cv] 7 Jun 2018 Deep supervision with additional labels for retinal vessel segmentation task Yishuo Zhang and Albert C.S. Chung Lo Kwee-Seong Medical Image Analysis Laboratory, Department of Computer Science and Engineering,

More information

DeepInsight: Multi-Task Multi-Scale Deep Learning for Mental Disorder Diagnosis

DeepInsight: Multi-Task Multi-Scale Deep Learning for Mental Disorder Diagnosis DING, HUO, HU, LU: DEEPINSIGHT 1 DeepInsight: Multi-Task Multi-Scale Deep Learning for Mental Disorder Diagnosis Mingyu Ding 1 d130143597@163.com Yuqi Huo 2 bnhony@163.com Jun Hu 2 junhu@ruc.edu.cn Zhiwu

More information

arxiv: v1 [cs.cv] 5 Dec 2014

arxiv: v1 [cs.cv] 5 Dec 2014 Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images Anh Nguyen University of Wyoming anguyen8@uwyo.edu Jason Yosinski Cornell University yosinski@cs.cornell.edu

More information

A HMM-based Pre-training Approach for Sequential Data

A HMM-based Pre-training Approach for Sequential Data A HMM-based Pre-training Approach for Sequential Data Luca Pasa 1, Alberto Testolin 2, Alessandro Sperduti 1 1- Department of Mathematics 2- Department of Developmental Psychology and Socialisation University

More information

Big Image-Omics Data Analytics for Clinical Outcome Prediction

Big Image-Omics Data Analytics for Clinical Outcome Prediction Big Image-Omics Data Analytics for Clinical Outcome Prediction Junzhou Huang, Ph.D. Associate Professor Dept. Computer Science & Engineering University of Texas at Arlington Dept. CSE, UT Arlington Scalable

More information

Tumor Cellularity Assessment. Rene Bidart

Tumor Cellularity Assessment. Rene Bidart Tumor Cellularity Assessment Rene Bidart 1 Goals Find location and class of all cell nuclei: Lymphocyte Cells Normal Epithelial Cells Malignant Epithelial Cells Why: Test effectiveness of pre-surgery treatment

More information

Early Diagnosis of Autism Disease by Multi-channel CNNs

Early Diagnosis of Autism Disease by Multi-channel CNNs Early Diagnosis of Autism Disease by Multi-channel CNNs Guannan Li 1,2, Mingxia Liu 2, Quansen Sun 1(&), Dinggang Shen 2(&), and Li Wang 2(&) 1 School of Computer Science and Engineering, Nanjing University

More information

Task 1: Machine Learning with Spike-Timing-Dependent Plasticity (STDP)

Task 1: Machine Learning with Spike-Timing-Dependent Plasticity (STDP) DARPA Report Task1 for Year 1 (Q1-Q4) Task 1: Machine Learning with Spike-Timing-Dependent Plasticity (STDP) 1. Shortcomings of the deep learning approach to artificial intelligence It has been established

More information

Comparison of Two Approaches for Direct Food Calorie Estimation

Comparison of Two Approaches for Direct Food Calorie Estimation Comparison of Two Approaches for Direct Food Calorie Estimation Takumi Ege and Keiji Yanai Department of Informatics, The University of Electro-Communications, Tokyo 1-5-1 Chofugaoka, Chofu-shi, Tokyo

More information

arxiv: v1 [cs.ne] 10 Sep 2018

arxiv: v1 [cs.ne] 10 Sep 2018 Fast and Efficient Information Transmission with Burst Spikes in Deep Spiking Neural Networks Seongsik Park, Seijoon Kim, Hyeokjun Choe, Sungroh Yoon Department of Electrical and Computer Engineering Seoul

More information

Do Deep Neural Networks Suffer from Crowding?

Do Deep Neural Networks Suffer from Crowding? Do Deep Neural Networks Suffer from Crowding? Anna Volokitin Gemma Roig ι Tomaso Poggio voanna@vision.ee.ethz.ch gemmar@mit.edu tp@csail.mit.edu Center for Brains, Minds and Machines, Massachusetts Institute

More information

Learning a Discriminative Filter Bank within a CNN for Fine-grained Recognition

Learning a Discriminative Filter Bank within a CNN for Fine-grained Recognition Learning a Discriminative Filter Bank within a CNN for Fine-grained Recognition Yaming Wang 1, Vlad I. Morariu 2, Larry S. Davis 1 1 University of Maryland, College Park 2 Adobe Research {wym, lsd}@umiacs.umd.edu

More information

Implementing a Robust Explanatory Bias in a Person Re-identification Network

Implementing a Robust Explanatory Bias in a Person Re-identification Network Implementing a Robust Explanatory Bias in a Person Re-identification Network Esube Bekele National Research Council Fellow Washington, DC esube.bekele.ctr@nrl.navy.mil Wallace E. Lawson Naval Research

More information

Deep Learning based FACS Action Unit Occurrence and Intensity Estimation

Deep Learning based FACS Action Unit Occurrence and Intensity Estimation Deep Learning based FACS Action Unit Occurrence and Intensity Estimation Amogh Gudi, H. Emrah Tasli, Tim M. den Uyl, Andreas Maroulis Vicarious Perception Technologies, Amsterdam, The Netherlands Abstract

More information

arxiv: v2 [cs.cv] 3 Jun 2018

arxiv: v2 [cs.cv] 3 Jun 2018 S4ND: Single-Shot Single-Scale Lung Nodule Detection Naji Khosravan and Ulas Bagci Center for Research in Computer Vision (CRCV), School of Computer Science, University of Central Florida, Orlando, FL.

More information

Interpreting Deep Neural Networks and their Predictions

Interpreting Deep Neural Networks and their Predictions Fraunhofer Image Processing Heinrich Hertz Institute Interpreting Deep Neural Networks and their Predictions Wojciech Samek ML Group, Fraunhofer HHI (joint work with S. Lapuschkin, A. Binder, G. Montavon,

More information

Training deep Autoencoders for collaborative filtering Oleksii Kuchaiev & Boris Ginsburg

Training deep Autoencoders for collaborative filtering Oleksii Kuchaiev & Boris Ginsburg Training deep Autoencoders for collaborative filtering Oleksii Kuchaiev & Boris Ginsburg Motivation Personalized recommendations 2 Key points (spoiler alert) 1. Deep autoencoder for collaborative filtering

More information

Deep Visual Attention Prediction

Deep Visual Attention Prediction IEEE TRANSACTIONS ON IMAGE PROCESSING 1 Deep Visual Attention Prediction Wenguan Wang, and Jianbing Shen, Senior Member, IEEE arxiv:1705.02544v3 [cs.cv] 22 Mar 2018 Abstract In this work, we aim to predict

More information

Discovery of Rare Phenotypes in Cellular Images Using Weakly Supervised Deep Learning

Discovery of Rare Phenotypes in Cellular Images Using Weakly Supervised Deep Learning Discovery of Rare Phenotypes in Cellular Images Using Weakly Supervised Deep Learning Heba Sailem *1, Mar Arias Garcia 2, Chris Bakal 2, Andrew Zisserman 1, and Jens Rittscher 1 1 Department of Engineering

More information

arxiv: v3 [cs.ne] 6 Jun 2016

arxiv: v3 [cs.ne] 6 Jun 2016 Synthesizing the preferred inputs for neurons in neural networks via deep generator networks Anh Nguyen anguyen8@uwyo.edu Alexey Dosovitskiy dosovits@cs.uni-freiburg.de arxiv:1605.09304v3 [cs.ne] 6 Jun

More information

Simultaneous Estimation of Food Categories and Calories with Multi-task CNN

Simultaneous Estimation of Food Categories and Calories with Multi-task CNN Simultaneous Estimation of Food Categories and Calories with Multi-task CNN Takumi Ege and Keiji Yanai The University of Electro-Communications, Tokyo 1 Introduction (1) Spread of meal management applications.

More information

Viewpoint Invariant Convolutional Networks for Identifying Risky Hand Hygiene Scenarios

Viewpoint Invariant Convolutional Networks for Identifying Risky Hand Hygiene Scenarios Viewpoint Invariant Convolutional Networks for Identifying Risky Hand Hygiene Scenarios Michelle Guo 1 Albert Haque 1 Serena Yeung 1 Jeffrey Jopling 1 Lance Downing 1 Alexandre Alahi 1,2 Brandi Campbell

More information

A Deep Multi-Level Network for Saliency Prediction

A Deep Multi-Level Network for Saliency Prediction A Deep Multi-Level Network for Saliency Prediction Marcella Cornia, Lorenzo Baraldi, Giuseppe Serra and Rita Cucchiara Dipartimento di Ingegneria Enzo Ferrari Università degli Studi di Modena e Reggio

More information

Expert identification of visual primitives used by CNNs during mammogram classification

Expert identification of visual primitives used by CNNs during mammogram classification Expert identification of visual primitives used by CNNs during mammogram classification Jimmy Wu a, Diondra Peck b, Scott Hsieh c, Vandana Dialani, MD d, Constance D. Lehman, MD e, Bolei Zhou a, Vasilis

More information

arxiv: v1 [cs.cv] 1 Feb 2017

arxiv: v1 [cs.cv] 1 Feb 2017 Visual Saliency Prediction Using a Mixture of Deep Neural Networks Samuel Dodge and Lina Karam Arizona State University {sfdodge,karam}@asu.edu arxiv:1702.00372v1 [cs.cv] 1 Feb 2017 Abstract Visual saliency

More information

AN EFFICIENT DIGITAL SUPPORT SYSTEM FOR DIAGNOSING BRAIN TUMOR

AN EFFICIENT DIGITAL SUPPORT SYSTEM FOR DIAGNOSING BRAIN TUMOR AN EFFICIENT DIGITAL SUPPORT SYSTEM FOR DIAGNOSING BRAIN TUMOR Yatendra Kashyap Corporate institute of science & Technology, Bhopal ---------------------------------------------------------------------***---------------------------------------------------------------------

More information

MAMMOGRAM AND TOMOSYNTHESIS CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS

MAMMOGRAM AND TOMOSYNTHESIS CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS University of Kentucky UKnowledge Theses and Dissertations--Computer Science Computer Science 2017 MAMMOGRAM AND TOMOSYNTHESIS CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS Xiaofei Zhang University

More information

arxiv: v1 [cs.cv] 13 Mar 2018

arxiv: v1 [cs.cv] 13 Mar 2018 RESOURCE AWARE DESIGN OF A DEEP CONVOLUTIONAL-RECURRENT NEURAL NETWORK FOR SPEECH RECOGNITION THROUGH AUDIO-VISUAL SENSOR FUSION Matthijs Van keirsbilck Bert Moons Marian Verhelst MICAS, Department of

More information

Finding the Unknown: Novelty Detection with Extreme Value Signatures of Deep Neural Activations

Finding the Unknown: Novelty Detection with Extreme Value Signatures of Deep Neural Activations A. Schultheiss, C. Käding, A. Freytag, J. Denzler Finding the Unknown: Novelty Detection with Extreme Value Signatures of Deep Neural Activations GCPR 2017 c Copyright by Springer. The final publication

More information

Deep CNNs for Diabetic Retinopathy Detection

Deep CNNs for Diabetic Retinopathy Detection Deep CNNs for Diabetic Retinopathy Detection Alex Tamkin Stanford University atamkin@stanford.edu Iain Usiri Stanford University iusiri@stanford.edu Chala Fufa Stanford University cfufa@stanford.edu 1

More information

Shu Kong. Department of Computer Science, UC Irvine

Shu Kong. Department of Computer Science, UC Irvine Ubiquitous Fine-Grained Computer Vision Shu Kong Department of Computer Science, UC Irvine Outline 1. Problem definition 2. Instantiation 3. Challenge and philosophy 4. Fine-grained classification with

More information

Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition

Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition Charles F. Cadieu, Ha Hong, Daniel L. K. Yamins, Nicolas Pinto, Diego Ardila, Ethan A. Solomon, Najib

More information