SlideShare ist ein Scribd-Unternehmen logo
1 von 25
Downloaden Sie, um offline zu lesen
Mix Conv: Mixed Depthwise
Convolutional Kernels
Hwang seung hyun
Yonsei University Severance Hospital CCIDS
Google Brain | BMVC 2019
2020.03.22
Introduction Related Work Methods and
Experiments
01 02 03
Conclusion
04
Yonsei Unversity Severance Hospital CCIDS
Contents
Mix Net
Introduction
• Recent trend in CNNs is to improve both accuracy and
efficiency
• Depthwise convolution are becoming more popular
[MobileNets, ShuffleNets, NASNet, AmoebaNet, MnasNet, EfficientNet]
• Mix Net focuses on kernel size.
• Recent studies showed large size kernels such as 5x5, and 7x7
kernels can potentially improve model accuracy and efficiency
with more details at the cost of more parameters and
computations.
• But do the always improve accuracy?
Introduction / Related Work / Methods and Experiments / Conclusion
Mix Net
Introduction
Introduction / Related Work / Methods and Experiments / Conclusion
Mix Net
Introduction
• Very large kernel sizes can hurt both accuracy and efficiency
• MixNet suggest limitation of single kernel size
• Propose Mixed Depthwise Convolution(MixConv)
→ Mixes up different kernel sizes in a single convolution operation, to
capture different patterns with various resolutions
• Partitions channels into multiple groups and apply different
kernel sizes to each group of channels
Introduction / Background / Methods and Experiments/ Conclusion
Mix Net
Introduction
Introduction / Related Work / Methods and Experiments / Conclusion
Mix Net
Introduction
• Significantly improve MobileNets accuracy and efficiency on
both ImageNet classification and COCO object detection
• Leverage Neural Architecture Search(NAS) to develop new
family of models named MixNets
• MixNet model achieved SOTA with 78.9% accuracy on
ImageNet dataset.
Introduction / Related Work / Methods and Experiments / Conclusion
Related Work
Efficient ConvNets
Introduction / Related Work / Methods and Experiments / Conclusion
• Depthwise Convolution has been increasingly popular in all
mobile-size Conv nets
• EfficientNet achieved SOTA accuracy on ImageNet, by using
depthwise and pointwise convolutions
Related Work
Multi-Scale Networks and Features
Introduction / Related Work / Methods and Experiments / Conclusion
• Use multiple branches in each layer to utilize different
operations in a single layer
• Inceptions, Inception-ResNet, ResNext, NASnet
* Part of Inception-resnetV2
Related Work
Neural Architecture Search
Introduction / Related Work / Methods and Experiments / Conclusion
• NAS has achieved better performance than hand-crafted
models by automating the design process and learning better
design choises
• Developed a new family of MixNets by adding MixConv into the
search space.
Methods and Experiments
MixConv Feature Map
Introduction / Related Work / Methodsand Experiments / Conclusion
• MixConv partitions channels into groups and applies different
kernel sizes to each group
1. Input tensor partitioned into “g” groups of tensors
2. Convolutional kernels grouped into ‘g’ virtual kernels
3. Virtual output 4. Final output
Methods and Experiments
MixConv Feature Map
Introduction / Related Work / Methodsand Experiments / Conclusion
* Grouped Convolution
Methods and Experiments
MixConv Design Choices
Introduction / Related Work / Methodsand Experiments / Conclusion
• Group Size g
→ With the help of NAS, authors experimented variety of group sizes from 1 to 5.
• Kernel Size per Group
→ Starts from 3x3, monotonically increases by 2 per group
• Channel Size per Group
→ Equal partition
→ Exponential Partition (i-th group with 2-i portion of total channels)
ex) 4 Group MixConv with total filter size 32
= Divide channels into (8,8,8,8) with equal partition
= Divide channels into (16, 8, 4, 4) with exponential partition
• Dilated Convolution
→ Since large kernels need more parameters and computations, an alternative is to use dilated
convolution. However, dilated convolutions usually have inferior accuracy than large kernel sizes.
Methods and Experiments
MixConv Performance on Mobile Nets – ImageNet
Introduction / Related Work / Methodsand Experiments / Conclusion
• Based on MobileNet V1, and V2, authors replaced all original
3x3 depthwise convolutional kernels with larger kernels of
MixCov kernels.
• MixConv generally uses much less parameters and FLOPS, but
its accuracy is similar or better than original.
• MixConv is much less sensitive to very large kernels
Methods and Experiments
MixConv Performance on Mobile Nets – ImageNet
Introduction / Related Work / Methodsand Experiments / Conclusion
Methods and Experiments
MixConv Performance on Mobile Nets – COCO object detection
Introduction / Related Work / Methodsand Experiments / Conclusion
• MixConv consistently achieves better efficiency and accuracy
than original
• 0.6% higher mAP(Mean Average Precision) on MobileNetV1
• 1.1% higher mAP on MobileNetV2 using fewer parameters
and FLOPS
Methods and Experiments
MixConv Performance on Mobile Nets – COCO object detection
Introduction / Related Work / Methodsand Experiments / Conclusion
Methods and Experiments
Ablation Study
Introduction / Related Work / Methodsand Experiments / Conclusion
1. MixConv for Single Layer
• In addition of applying MixConv to the whole network, analyzed per-
layer performance on MobileNetV2.
• MixConv achieved similar of slightly better performance for most of
the layers
Methods and Experiments
Ablation Study
Introduction / Related Work / Methodsand Experiments / Conclusion
2. Channel Partition Methods
• Equal Partition vs. Exponential Partition
• Exponential partition required less parameters and FLOPS for the same
kernel size, by assigning more channels to smaller kernels.
• No clear winner between two partition methods
Methods and Experiments
Ablation Study
Introduction / Related Work / Methodsand Experiments / Conclusion
3. Dilated Convolution
• Dilated convolution has reasonable performance for small kernels, but
the accuracy drops quickly for large kernels.
• Dilated Convolution skips a lot of local information within large kernels,
dropping accuracy
MixNet – Architecture Search
Methods and Experiments
Introduction / Related Work / Methodsand Experiments / Conclusion
• Leverage recent progress in neural architecture search to
develop a new family of MixConv-based models, named as
MixNets.
• NAS search settings are similar to recent MnasNet
• Used MobileNetV2 as baseline network
• NAS search for the best kernel size, expansion ratio, channel
size, etc
• Equal channel partition
• No dilated convolutions
MixNet – Architecture Search
Methods and Experiments
Introduction / Related Work / Methodsand Experiments / Conclusion
• New additions to search space
* Swish Activation
* Squeeze-and-Excitation Module
* Grouped Convolutions with group size from 1 to 5
* Adopted MixConv as the basic convolutional operation
* Swish Activation
MixNet – Architectures
Methods and Experiments
Introduction / Related Work / Methodsand Experiments / Conclusion
• Small Kernels are more common in early stage for saving computational
cost
• Large Kernels are more common in later stage for better accuracy
• MixNets are capable of utilizing very large kernels such as 9x9 and
11x11 to capture high-resolution patterns from input images
MixNet – Performance on ImageNet
Methods and Experiments
Introduction / Related Work / Methodsand Experiments / Conclusion
• Obtained MixNet-S and M from NAS, and scaled up MixNet-M with depth
multiplier 1.3 to obtain MixNet-L
• MixNets outperform all latest mobile ConvNets
Conclusion
Introduction / Related Work / Methods and Experiments / Conclusion
• Studied impact of kernel size for depthwise convolution
• Traditional depthwise convolution suffers from the limitations of single
kernel size
• Proposed MixConv which mixes multiple kernels in a single op
• Improves the accuracy and efficiency for MobileNets on both Image
Classification and Object Detection tasks
• Further Developed a new family of MixNets using NAS
• MixNets achieved significantly better accuracy and efficiency than all latest
mobile ConvNets

Weitere ähnliche Inhalte

Was ist angesagt?

PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...
PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...
PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...Sunghoon Joo
 
Mlp mixer image_process_210613 deeplearning paper review!
Mlp mixer image_process_210613 deeplearning paper review!Mlp mixer image_process_210613 deeplearning paper review!
Mlp mixer image_process_210613 deeplearning paper review!taeseon ryu
 
PR095: Modularity Matters: Learning Invariant Relational Reasoning Tasks
PR095: Modularity Matters: Learning Invariant Relational Reasoning TasksPR095: Modularity Matters: Learning Invariant Relational Reasoning Tasks
PR095: Modularity Matters: Learning Invariant Relational Reasoning TasksJinwon Lee
 
Finding the best solution for Image Processing
Finding the best solution for Image ProcessingFinding the best solution for Image Processing
Finding the best solution for Image ProcessingTech Triveni
 
Distributed Mininet with Symbiosis
Distributed Mininet with SymbiosisDistributed Mininet with Symbiosis
Distributed Mininet with SymbiosisJason Liu
 
Exploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation LearningExploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation LearningSungchul Kim
 
Large-scale Video Classification with Convolutional Neural Network
Large-scale Video Classification with Convolutional Neural NetworkLarge-scale Video Classification with Convolutional Neural Network
Large-scale Video Classification with Convolutional Neural NetworkKhalidKhan412
 
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.Sunghoon Joo
 
PR12-193 NISP: Pruning Networks using Neural Importance Score Propagation
PR12-193 NISP: Pruning Networks using Neural Importance Score PropagationPR12-193 NISP: Pruning Networks using Neural Importance Score Propagation
PR12-193 NISP: Pruning Networks using Neural Importance Score PropagationTaesu Kim
 
PR-351: Adaptive Aggregation Networks for Class-Incremental Learning
PR-351: Adaptive Aggregation Networks for Class-Incremental LearningPR-351: Adaptive Aggregation Networks for Class-Incremental Learning
PR-351: Adaptive Aggregation Networks for Class-Incremental LearningSunghoon Joo
 
Deep Learning Fast MRI Using Channel Attention in Magnitude Domain
Deep Learning Fast MRI Using Channel Attention in Magnitude DomainDeep Learning Fast MRI Using Channel Attention in Magnitude Domain
Deep Learning Fast MRI Using Channel Attention in Magnitude DomainJoonhyung Lee
 
ELLA LC algorithm presentation in ICIP 2016
ELLA LC algorithm presentation in ICIP 2016ELLA LC algorithm presentation in ICIP 2016
ELLA LC algorithm presentation in ICIP 2016InVID Project
 
Mlp mixer an all-mlp architecture for vision
Mlp mixer  an all-mlp architecture for visionMlp mixer  an all-mlp architecture for vision
Mlp mixer an all-mlp architecture for visionJaey Jeong
 
Pruning convolutional neural networks for resource efficient inference
Pruning convolutional neural networks for resource efficient inferencePruning convolutional neural networks for resource efficient inference
Pruning convolutional neural networks for resource efficient inferenceKaushalya Madhawa
 
Fully convolutional networks for semantic segmentation
Fully convolutional networks for semantic segmentation Fully convolutional networks for semantic segmentation
Fully convolutional networks for semantic segmentation heedaeKwon
 
Paraphrasing complex network
Paraphrasing complex networkParaphrasing complex network
Paraphrasing complex networkNAVER Engineering
 

Was ist angesagt? (20)

PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...
PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...
PR-187 : MorphNet: Fast & Simple Resource-Constrained Structure Learning of D...
 
Mlp mixer image_process_210613 deeplearning paper review!
Mlp mixer image_process_210613 deeplearning paper review!Mlp mixer image_process_210613 deeplearning paper review!
Mlp mixer image_process_210613 deeplearning paper review!
 
poster
posterposter
poster
 
PR095: Modularity Matters: Learning Invariant Relational Reasoning Tasks
PR095: Modularity Matters: Learning Invariant Relational Reasoning TasksPR095: Modularity Matters: Learning Invariant Relational Reasoning Tasks
PR095: Modularity Matters: Learning Invariant Relational Reasoning Tasks
 
Dl
DlDl
Dl
 
Finding the best solution for Image Processing
Finding the best solution for Image ProcessingFinding the best solution for Image Processing
Finding the best solution for Image Processing
 
EfficientNet
EfficientNetEfficientNet
EfficientNet
 
Distributed Mininet with Symbiosis
Distributed Mininet with SymbiosisDistributed Mininet with Symbiosis
Distributed Mininet with Symbiosis
 
Exploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation LearningExploring Simple Siamese Representation Learning
Exploring Simple Siamese Representation Learning
 
Large-scale Video Classification with Convolutional Neural Network
Large-scale Video Classification with Convolutional Neural NetworkLarge-scale Video Classification with Convolutional Neural Network
Large-scale Video Classification with Convolutional Neural Network
 
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.
PR-373: Revisiting ResNets: Improved Training and Scaling Strategies.
 
PR12-193 NISP: Pruning Networks using Neural Importance Score Propagation
PR12-193 NISP: Pruning Networks using Neural Importance Score PropagationPR12-193 NISP: Pruning Networks using Neural Importance Score Propagation
PR12-193 NISP: Pruning Networks using Neural Importance Score Propagation
 
PR-351: Adaptive Aggregation Networks for Class-Incremental Learning
PR-351: Adaptive Aggregation Networks for Class-Incremental LearningPR-351: Adaptive Aggregation Networks for Class-Incremental Learning
PR-351: Adaptive Aggregation Networks for Class-Incremental Learning
 
Deep Learning Fast MRI Using Channel Attention in Magnitude Domain
Deep Learning Fast MRI Using Channel Attention in Magnitude DomainDeep Learning Fast MRI Using Channel Attention in Magnitude Domain
Deep Learning Fast MRI Using Channel Attention in Magnitude Domain
 
ELLA LC algorithm presentation in ICIP 2016
ELLA LC algorithm presentation in ICIP 2016ELLA LC algorithm presentation in ICIP 2016
ELLA LC algorithm presentation in ICIP 2016
 
Mlp mixer an all-mlp architecture for vision
Mlp mixer  an all-mlp architecture for visionMlp mixer  an all-mlp architecture for vision
Mlp mixer an all-mlp architecture for vision
 
Pruning convolutional neural networks for resource efficient inference
Pruning convolutional neural networks for resource efficient inferencePruning convolutional neural networks for resource efficient inference
Pruning convolutional neural networks for resource efficient inference
 
Fully convolutional networks for semantic segmentation
Fully convolutional networks for semantic segmentation Fully convolutional networks for semantic segmentation
Fully convolutional networks for semantic segmentation
 
Paraphrasing complex network
Paraphrasing complex networkParaphrasing complex network
Paraphrasing complex network
 
Network recasting
Network recastingNetwork recasting
Network recasting
 

Ähnlich wie Mix Conv: Mixed Depthwise Convolutional Kernels

ResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksSeunghyun Hwang
 
ConvNeXt.pptx
ConvNeXt.pptxConvNeXt.pptx
ConvNeXt.pptxYanhuaSi
 
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptx
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptxEfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptx
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptxssuser2624f71
 
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural NetworksPR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural NetworksJinwon Lee
 
PR-366: A ConvNet for 2020s
PR-366: A ConvNet for 2020sPR-366: A ConvNet for 2020s
PR-366: A ConvNet for 2020sJinwon Lee
 
Group Communication Techniques in Overlay Networks
Group Communication Techniques in Overlay NetworksGroup Communication Techniques in Overlay Networks
Group Communication Techniques in Overlay NetworksKnut-Helge Vik
 
04 Deep CNN (Ch_01 to Ch_3).pptx
04 Deep CNN (Ch_01 to Ch_3).pptx04 Deep CNN (Ch_01 to Ch_3).pptx
04 Deep CNN (Ch_01 to Ch_3).pptxZainULABIDIN496386
 
ConvNeXt: A ConvNet for the 2020s explained
ConvNeXt: A ConvNet for the 2020s explainedConvNeXt: A ConvNet for the 2020s explained
ConvNeXt: A ConvNet for the 2020s explainedSushant Gautam
 
PR243: Designing Network Design Spaces
PR243: Designing Network Design SpacesPR243: Designing Network Design Spaces
PR243: Designing Network Design SpacesJinwon Lee
 
Learning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutLearning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutSeunghyun Hwang
 
Convolutional Neural Networks : Popular Architectures
Convolutional Neural Networks : Popular ArchitecturesConvolutional Neural Networks : Popular Architectures
Convolutional Neural Networks : Popular Architecturesananth
 
Efficient de cvpr_2020_paper
Efficient de cvpr_2020_paperEfficient de cvpr_2020_paper
Efficient de cvpr_2020_papershanullah3
 
Cvpr 2018 papers review (efficient computing)
Cvpr 2018 papers review (efficient computing)Cvpr 2018 papers review (efficient computing)
Cvpr 2018 papers review (efficient computing)DonghyunKang12
 
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video StreamingOn Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video StreamingAlpen-Adria-Universität
 
Handwritten Digit Recognition and performance of various modelsation[autosaved]
Handwritten Digit Recognition and performance of various modelsation[autosaved]Handwritten Digit Recognition and performance of various modelsation[autosaved]
Handwritten Digit Recognition and performance of various modelsation[autosaved]SubhradeepMaji
 
Machine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural NetworkMachine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural NetworkRichard Kuo
 
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...Jinwon Lee
 
Comparison of Learning Algorithms for Handwritten Digit Recognition
Comparison of Learning Algorithms for Handwritten Digit RecognitionComparison of Learning Algorithms for Handwritten Digit Recognition
Comparison of Learning Algorithms for Handwritten Digit RecognitionSafaa Alnabulsi
 

Ähnlich wie Mix Conv: Mixed Depthwise Convolutional Kernels (20)

ResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksResNeSt: Split-Attention Networks
ResNeSt: Split-Attention Networks
 
ConvNeXt.pptx
ConvNeXt.pptxConvNeXt.pptx
ConvNeXt.pptx
 
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptx
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptxEfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptx
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.pptx
 
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural NetworksPR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
PR-169: EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
 
PR-366: A ConvNet for 2020s
PR-366: A ConvNet for 2020sPR-366: A ConvNet for 2020s
PR-366: A ConvNet for 2020s
 
Group Communication Techniques in Overlay Networks
Group Communication Techniques in Overlay NetworksGroup Communication Techniques in Overlay Networks
Group Communication Techniques in Overlay Networks
 
04 Deep CNN (Ch_01 to Ch_3).pptx
04 Deep CNN (Ch_01 to Ch_3).pptx04 Deep CNN (Ch_01 to Ch_3).pptx
04 Deep CNN (Ch_01 to Ch_3).pptx
 
ConvNeXt: A ConvNet for the 2020s explained
ConvNeXt: A ConvNet for the 2020s explainedConvNeXt: A ConvNet for the 2020s explained
ConvNeXt: A ConvNet for the 2020s explained
 
PR243: Designing Network Design Spaces
PR243: Designing Network Design SpacesPR243: Designing Network Design Spaces
PR243: Designing Network Design Spaces
 
Learning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutLearning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted Dropout
 
Convolutional Neural Networks : Popular Architectures
Convolutional Neural Networks : Popular ArchitecturesConvolutional Neural Networks : Popular Architectures
Convolutional Neural Networks : Popular Architectures
 
Efficient de cvpr_2020_paper
Efficient de cvpr_2020_paperEfficient de cvpr_2020_paper
Efficient de cvpr_2020_paper
 
Cvpr 2018 papers review (efficient computing)
Cvpr 2018 papers review (efficient computing)Cvpr 2018 papers review (efficient computing)
Cvpr 2018 papers review (efficient computing)
 
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video StreamingOn Optimizing Resource Utilization in AVC-based Real-time Video Streaming
On Optimizing Resource Utilization in AVC-based Real-time Video Streaming
 
Handwritten Digit Recognition and performance of various modelsation[autosaved]
Handwritten Digit Recognition and performance of various modelsation[autosaved]Handwritten Digit Recognition and performance of various modelsation[autosaved]
Handwritten Digit Recognition and performance of various modelsation[autosaved]
 
1801.06434
1801.064341801.06434
1801.06434
 
Machine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural NetworkMachine Learning - Convolutional Neural Network
Machine Learning - Convolutional Neural Network
 
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...
PR-344: A Battle of Network Structures: An Empirical Study of CNN, Transforme...
 
Comparison of Learning Algorithms for Handwritten Digit Recognition
Comparison of Learning Algorithms for Handwritten Digit RecognitionComparison of Learning Algorithms for Handwritten Digit Recognition
Comparison of Learning Algorithms for Handwritten Digit Recognition
 
Conformer review
Conformer reviewConformer review
Conformer review
 

Mehr von Seunghyun Hwang

An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...Seunghyun Hwang
 
Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Seunghyun Hwang
 
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Seunghyun Hwang
 
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Seunghyun Hwang
 
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Seunghyun Hwang
 
End-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersEnd-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersSeunghyun Hwang
 
Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Seunghyun Hwang
 
Segmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSegmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSeunghyun Hwang
 
Progressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsProgressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsSeunghyun Hwang
 
A Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsA Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsSeunghyun Hwang
 
How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?Seunghyun Hwang
 
DeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementDeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementSeunghyun Hwang
 
Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Seunghyun Hwang
 
A Probabilistic U-Net for Segmentation of Ambiguous Images
A Probabilistic U-Net for Segmentation of Ambiguous ImagesA Probabilistic U-Net for Segmentation of Ambiguous Images
A Probabilistic U-Net for Segmentation of Ambiguous ImagesSeunghyun Hwang
 
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...Seunghyun Hwang
 
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisLarge Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisSeunghyun Hwang
 

Mehr von Seunghyun Hwang (16)

An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...
 
Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...
 
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
 
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
 
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
 
End-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersEnd-to-End Object Detection with Transformers
End-to-End Object Detection with Transformers
 
Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation
 
Segmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSegmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding Cell
 
Progressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsProgressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representations
 
A Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsA Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual Representations
 
How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?
 
DeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementDeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary Refinement
 
Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...
 
A Probabilistic U-Net for Segmentation of Ambiguous Images
A Probabilistic U-Net for Segmentation of Ambiguous ImagesA Probabilistic U-Net for Segmentation of Ambiguous Images
A Probabilistic U-Net for Segmentation of Ambiguous Images
 
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
 
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisLarge Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image Synthesis
 

Kürzlich hochgeladen

Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Scriptwesley chun
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...Neo4j
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking MenDelhi Call girls
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEarley Information Science
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?Igalia
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking MenDelhi Call girls
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsJoaquim Jorge
 

Kürzlich hochgeladen (20)

Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Script
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
 
Artificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and MythsArtificial Intelligence: Facts and Myths
Artificial Intelligence: Facts and Myths
 

Mix Conv: Mixed Depthwise Convolutional Kernels

  • 1. Mix Conv: Mixed Depthwise Convolutional Kernels Hwang seung hyun Yonsei University Severance Hospital CCIDS Google Brain | BMVC 2019 2020.03.22
  • 2. Introduction Related Work Methods and Experiments 01 02 03 Conclusion 04 Yonsei Unversity Severance Hospital CCIDS Contents
  • 3. Mix Net Introduction • Recent trend in CNNs is to improve both accuracy and efficiency • Depthwise convolution are becoming more popular [MobileNets, ShuffleNets, NASNet, AmoebaNet, MnasNet, EfficientNet] • Mix Net focuses on kernel size. • Recent studies showed large size kernels such as 5x5, and 7x7 kernels can potentially improve model accuracy and efficiency with more details at the cost of more parameters and computations. • But do the always improve accuracy? Introduction / Related Work / Methods and Experiments / Conclusion
  • 4. Mix Net Introduction Introduction / Related Work / Methods and Experiments / Conclusion
  • 5. Mix Net Introduction • Very large kernel sizes can hurt both accuracy and efficiency • MixNet suggest limitation of single kernel size • Propose Mixed Depthwise Convolution(MixConv) → Mixes up different kernel sizes in a single convolution operation, to capture different patterns with various resolutions • Partitions channels into multiple groups and apply different kernel sizes to each group of channels Introduction / Background / Methods and Experiments/ Conclusion
  • 6. Mix Net Introduction Introduction / Related Work / Methods and Experiments / Conclusion
  • 7. Mix Net Introduction • Significantly improve MobileNets accuracy and efficiency on both ImageNet classification and COCO object detection • Leverage Neural Architecture Search(NAS) to develop new family of models named MixNets • MixNet model achieved SOTA with 78.9% accuracy on ImageNet dataset. Introduction / Related Work / Methods and Experiments / Conclusion
  • 8. Related Work Efficient ConvNets Introduction / Related Work / Methods and Experiments / Conclusion • Depthwise Convolution has been increasingly popular in all mobile-size Conv nets • EfficientNet achieved SOTA accuracy on ImageNet, by using depthwise and pointwise convolutions
  • 9. Related Work Multi-Scale Networks and Features Introduction / Related Work / Methods and Experiments / Conclusion • Use multiple branches in each layer to utilize different operations in a single layer • Inceptions, Inception-ResNet, ResNext, NASnet * Part of Inception-resnetV2
  • 10. Related Work Neural Architecture Search Introduction / Related Work / Methods and Experiments / Conclusion • NAS has achieved better performance than hand-crafted models by automating the design process and learning better design choises • Developed a new family of MixNets by adding MixConv into the search space.
  • 11. Methods and Experiments MixConv Feature Map Introduction / Related Work / Methodsand Experiments / Conclusion • MixConv partitions channels into groups and applies different kernel sizes to each group 1. Input tensor partitioned into “g” groups of tensors 2. Convolutional kernels grouped into ‘g’ virtual kernels 3. Virtual output 4. Final output
  • 12. Methods and Experiments MixConv Feature Map Introduction / Related Work / Methodsand Experiments / Conclusion * Grouped Convolution
  • 13. Methods and Experiments MixConv Design Choices Introduction / Related Work / Methodsand Experiments / Conclusion • Group Size g → With the help of NAS, authors experimented variety of group sizes from 1 to 5. • Kernel Size per Group → Starts from 3x3, monotonically increases by 2 per group • Channel Size per Group → Equal partition → Exponential Partition (i-th group with 2-i portion of total channels) ex) 4 Group MixConv with total filter size 32 = Divide channels into (8,8,8,8) with equal partition = Divide channels into (16, 8, 4, 4) with exponential partition • Dilated Convolution → Since large kernels need more parameters and computations, an alternative is to use dilated convolution. However, dilated convolutions usually have inferior accuracy than large kernel sizes.
  • 14. Methods and Experiments MixConv Performance on Mobile Nets – ImageNet Introduction / Related Work / Methodsand Experiments / Conclusion • Based on MobileNet V1, and V2, authors replaced all original 3x3 depthwise convolutional kernels with larger kernels of MixCov kernels. • MixConv generally uses much less parameters and FLOPS, but its accuracy is similar or better than original. • MixConv is much less sensitive to very large kernels
  • 15. Methods and Experiments MixConv Performance on Mobile Nets – ImageNet Introduction / Related Work / Methodsand Experiments / Conclusion
  • 16. Methods and Experiments MixConv Performance on Mobile Nets – COCO object detection Introduction / Related Work / Methodsand Experiments / Conclusion • MixConv consistently achieves better efficiency and accuracy than original • 0.6% higher mAP(Mean Average Precision) on MobileNetV1 • 1.1% higher mAP on MobileNetV2 using fewer parameters and FLOPS
  • 17. Methods and Experiments MixConv Performance on Mobile Nets – COCO object detection Introduction / Related Work / Methodsand Experiments / Conclusion
  • 18. Methods and Experiments Ablation Study Introduction / Related Work / Methodsand Experiments / Conclusion 1. MixConv for Single Layer • In addition of applying MixConv to the whole network, analyzed per- layer performance on MobileNetV2. • MixConv achieved similar of slightly better performance for most of the layers
  • 19. Methods and Experiments Ablation Study Introduction / Related Work / Methodsand Experiments / Conclusion 2. Channel Partition Methods • Equal Partition vs. Exponential Partition • Exponential partition required less parameters and FLOPS for the same kernel size, by assigning more channels to smaller kernels. • No clear winner between two partition methods
  • 20. Methods and Experiments Ablation Study Introduction / Related Work / Methodsand Experiments / Conclusion 3. Dilated Convolution • Dilated convolution has reasonable performance for small kernels, but the accuracy drops quickly for large kernels. • Dilated Convolution skips a lot of local information within large kernels, dropping accuracy
  • 21. MixNet – Architecture Search Methods and Experiments Introduction / Related Work / Methodsand Experiments / Conclusion • Leverage recent progress in neural architecture search to develop a new family of MixConv-based models, named as MixNets. • NAS search settings are similar to recent MnasNet • Used MobileNetV2 as baseline network • NAS search for the best kernel size, expansion ratio, channel size, etc • Equal channel partition • No dilated convolutions
  • 22. MixNet – Architecture Search Methods and Experiments Introduction / Related Work / Methodsand Experiments / Conclusion • New additions to search space * Swish Activation * Squeeze-and-Excitation Module * Grouped Convolutions with group size from 1 to 5 * Adopted MixConv as the basic convolutional operation * Swish Activation
  • 23. MixNet – Architectures Methods and Experiments Introduction / Related Work / Methodsand Experiments / Conclusion • Small Kernels are more common in early stage for saving computational cost • Large Kernels are more common in later stage for better accuracy • MixNets are capable of utilizing very large kernels such as 9x9 and 11x11 to capture high-resolution patterns from input images
  • 24. MixNet – Performance on ImageNet Methods and Experiments Introduction / Related Work / Methodsand Experiments / Conclusion • Obtained MixNet-S and M from NAS, and scaled up MixNet-M with depth multiplier 1.3 to obtain MixNet-L • MixNets outperform all latest mobile ConvNets
  • 25. Conclusion Introduction / Related Work / Methods and Experiments / Conclusion • Studied impact of kernel size for depthwise convolution • Traditional depthwise convolution suffers from the limitations of single kernel size • Proposed MixConv which mixes multiple kernels in a single op • Improves the accuracy and efficiency for MobileNets on both Image Classification and Object Detection tasks • Further Developed a new family of MixNets using NAS • MixNets achieved significantly better accuracy and efficiency than all latest mobile ConvNets