SlideShare a Scribd company logo
1 of 29
Download to read offline
Optical flow estimation
with Blueoil
K-Inoue @ki42 & Oscar @wang
Blueoil Division
1
What is optical flow estimation?
● Optical flow...
represents the apparent motion of objects.
● Optical flow estimation...
can predict the movement of objects in a video.
Miloud, Hadj achour. (2017). Fragmentation de métal liquide dans l’eau.
https://www.codeproject.com/Articles/1192205/Capturing-motion-from-video-using-the-Emgu-CV-libr
2
Optical flow estimation is important
● Widely used by insects and birds
● Practical usage
○ Analyze motion
○ Avoid collision
○ Assist in navigation
● Real-world Applications
○ Video/Motion classification
○ Navigation assistance
■ Self driving cars
■ Drones https://nanonets.com/blog/optical-flow/
3
DL approaches are increasing
4
● 2015. FlowNet S (Simple)
● 2015. FlowNet C (Correlated)
● 2016. FlowNet 2
● 2018. LiteFlowNet
● ...
Color coding
https://www.youtube.com/watch?v=k_wkDLJ8lJE https://www.youtube.com/watch?v=pfQ0zFwv-hM
However...
● Existing DL approaches require GPU to execute
👎 High power consumption
👎 Low runtime speed on CPU environment
● We propose LmFlowNet S 👍
○ Modification of FlowNet S [P. Fischer+ 2015]
○ Goals:
■ Edge Computing
■ Run on FPGA-based accelerator
■ Use quantization to reduce inference time
while achieving good prediction performance
5
FlowNet S vs LmFlowNet S
6
Network of FlowNet S
7
[N, 384, 512, 6]
[N, 384, 512, 2]
Detailed ops inside each color block is shown in appendix
Network of FlowNet S
8
[N, 384, 512, 6]
[N, 384, 512, 2]
Encoder
Decoder Detailed ops inside each color block is shown in appendix
Network of FlowNet S
9
[N, 384, 512, 6]
[N, 384, 512, 2]
Detailed ops inside each color block is shown in appendix
Not supported
by Blueoil DLK
10
Network of LmFlowNet S (DLK supported)
[N, 384, 512, 6]
[N, 384, 512, 2]
Detailed ops inside each color block is shown in appendix
11
Network of LmFlowNet S
[N, 384, 512, 6]
[N, 384, 512, 2]
Quantized
Detailed quantization inside each color block is shown in appendix
Loss function: End Point Error (EPE)
12
(x1
,y1
)
(x2
,y2
)
EPEflow2
EPEflow3
EPEflow4
EPEflow5
EPEflow6
Weighted EPE =
0.32 * EPEflow2
+
0.08 * EPEflow3
+
0.02 * EPEflow4
+
0.01 * EPEflow5
+
0.005 * EPEflow6
Down-
sampled
Ground
Truth
Training hyper-parameters are shown in appendix
24 x 32
48 x 6496 x 128
192 x 256
12 x 16
Artificial dataset: Flying Chairs
13
● Dataset
Name Frame pairs Train validation ratio size
Flying Chairs 22,872 9:1 30GB
● Data Augmentation
○ Crop, Rotate, Translate, FlipLeftRight, FlipTopBottom
○ Gaussian noise, Brightness, Contrast, Gamma, and Color
Parameters used in data augmentation are shown in appendix
https://arxiv.org/pdf/1504.06852.pdf
14
Results
15
Results - Avg. EPE & Inference time
Method
Avg. EPE (pixel)
(Flying Chairs)
Inference time per frame (ms) [1]
CPU (dlk-convert) GPU (tensorflow)
FlowNet S 2.94 - 11.65
LmFlowNet S 5.33 1360.49 13.81
LmFlowNet S
Quantized
9.01 637.467 17.60
[1] CPU and GPU specs available in appedix
FlowNet S
Trained for 1.2M
LmFlowNet S
Trained for 400K
LmFlowNet S Quantized
Trained for 400K
16
Live Demo
17
Live demonstration
● Three demonstration
○ FlowNet S
○ LmFlowNet S
○ LmFlowNet S Quantized
● NOTE: Running on GPU (not on CPU / FPGA)
○ Failed to run on CPU/FPGA due to several problems 😢
■ etc. segmentation fault, memory error...
○ Fixing and debugging them in the future 👊
18
Challenges
19
Challenges
● Training takes a very long time ( > 2 weeks...😢)
○ Heavy data augmentation & pre-processing
■ Pre-processing on GPU is not supported now.
● Unique network structure, not compatible with Blueoil
○ Input is a stack of 2 images (6 channels)
○ Multiple and branched outputs
● DLK Limitation. No documentation. 🤯
○ No support for kernel size 7x7, 5x5
○ No support for Conv2dTranspose
○ Cannot concat quantized value and float together
○ Requires the depth of Space2Depth to be 32 * N 20
Thank you for your
attention!
our source code:
https://github.com/ki-lm/blueoil/tree/lmflownets
21
22
Appendix
23
24
LmFlowNet S | Training & Hyper-params
● Optimizer: Adam
● Max steps: 1200k
● Fixed parameters in Adam: β1
=0.9 and β2
= 0.999.
● Batch size: 8
● Learning rate
○ values: [1e-4
, 5e-5
, 2.5e-5
,1.25e-5
,6.25e-6
]
○ boundaries: [400K, 600K, 800K, 1000K]
● Learning rate for quantization
○ values: [1.25e-5
, 1e-4
, 5e-5
, 2.5e-5
,1.25e-5
,6.25e-6
]
○ boundaries: [50K, 400K, 600K, 800K, 1000K]
25
Tokunaga Scheduling 😆
LmFlowNet S | Data Augmentation
● Translation: [20%, 20%] of the image width for x and y
● Rotation: [17o
, 17o
]
● Scaling: [0.9, 2.0]
● Gaussian noise: sigma uniformly sampled from [0, 0.04]
● Contrast: [0.8, 0.4]
● Multiplicative color changes to the RGB channels per
image: [0.5, 2]
● Gamma values: [0.7, 1.5]
● Additive brightness changes: Gaussian with a sigma of
0.2
26
27
CPU & GPU specs
● CPU
○ Intel(R) Core(TM) i7-5500U CPU @ 2.40GHz
○ 4 cores, 8 threads
○ Inference time (ms)
■ LmFlowNet S: 1358.39, 1358.39, 1293.26, 1307.28, 1485.11
■ LmFlowNet S Quant: 640.541, 640.732, 635.354, 635.354, 635.354
● GPU
○ NVIDIA Tesla V100 on DGX-1
○ Inference time (ms)
■ FlowNet S: 0.0111, 0.0105, 0.0116, 0.0117, 0.0119, 0.0118, 0.0121, 0.0118, 0.0116,
0.0124
■ LmFlowNet S: 0.0137, 0.0140, 0.0137, 0.0136, 0.0141, 0.0135, 0.0139, 0.0138, 0.0136,
0.0142
■ LmFlowNet S Quant: 0.0168, 0.0161, 0.0165, 0.0177, 0.0173, 0.0166, 0.0186, 0.0195,
0.0183, 0.0187
Our FlowNet S versions
28
Version Architecture DLK support
V1
(FlowNet S)
Same as the paper X
V2
7x7, 5x5 => 3x3
Striding 2 => SpaceToDepth
X
V3
Conv2dTranspose =>
ResizeNearestNeighbor + Conv2d
ResizeBilinear =>
ResizeNearestNeighbor
X
V3 Quant.
(LmFlowNet S)
Quantize except first, last layer,
and activation before last layer
△
V4 Quant.
Change all output depths from
SpaceToDepth to 32 * N
O
List of source code links
● FlowNet S/C, and 2 (TensorFlow):
https://github.com/sampepose/flownet2-tf/
● FlowNet S/C (Original paper, Caffe):
https://lmb.informatik.uni-freiburg.de/Publications/2015/DF
IB15/
● FlowNet 2 (Original paper, Caffe):
https://github.com/lmb-freiburg/flownet2
29

More Related Content

What's hot

Basic Generative Adversarial Networks
Basic Generative Adversarial NetworksBasic Generative Adversarial Networks
Basic Generative Adversarial NetworksDong Heon Cho
 
Chapter 4 Image Processing: Image Transformation
Chapter 4 Image Processing: Image TransformationChapter 4 Image Processing: Image Transformation
Chapter 4 Image Processing: Image TransformationVarun Ojha
 
Convolutional Neural Network Models - Deep Learning
Convolutional Neural Network Models - Deep LearningConvolutional Neural Network Models - Deep Learning
Convolutional Neural Network Models - Deep LearningMohamed Loey
 
Understanding neural radiance fields
Understanding neural radiance fieldsUnderstanding neural radiance fields
Understanding neural radiance fieldsVarun Bhaseen
 
Image segmentation
Image segmentationImage segmentation
Image segmentationKuppusamy P
 
6.frequency domain image_processing
6.frequency domain image_processing6.frequency domain image_processing
6.frequency domain image_processingNashid Alam
 
Object Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksObject Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksUsman Qayyum
 
Wavelet based image compression technique
Wavelet based image compression techniqueWavelet based image compression technique
Wavelet based image compression techniquePriyanka Pachori
 
Computer vision - images and image filtering
Computer vision - images and image filtering Computer vision - images and image filtering
Computer vision - images and image filtering Wael Badawy
 
Convolutional Neural Network
Convolutional Neural NetworkConvolutional Neural Network
Convolutional Neural NetworkVignesh Suresh
 
CNN and its applications by ketaki
CNN and its applications by ketakiCNN and its applications by ketaki
CNN and its applications by ketakiKetaki Patwari
 
Image segmentation
Image segmentationImage segmentation
Image segmentationRania H
 
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)Philip Hammer
 
Enhancement in frequency domain
Enhancement in frequency domainEnhancement in frequency domain
Enhancement in frequency domainAshish Kumar
 
Digital image processing img smoothning
Digital image processing img smoothningDigital image processing img smoothning
Digital image processing img smoothningVinay Gupta
 
Image processing, Noise, Noise Removal filters
Image processing, Noise, Noise Removal filtersImage processing, Noise, Noise Removal filters
Image processing, Noise, Noise Removal filtersKuppusamy P
 
Digital Image Processing: Digital Image Fundamentals
Digital Image Processing: Digital Image FundamentalsDigital Image Processing: Digital Image Fundamentals
Digital Image Processing: Digital Image FundamentalsMostafa G. M. Mostafa
 
Object detection - RCNNs vs Retinanet
Object detection - RCNNs vs RetinanetObject detection - RCNNs vs Retinanet
Object detection - RCNNs vs RetinanetRishabh Indoria
 

What's hot (20)

Basic Generative Adversarial Networks
Basic Generative Adversarial NetworksBasic Generative Adversarial Networks
Basic Generative Adversarial Networks
 
Chapter 4 Image Processing: Image Transformation
Chapter 4 Image Processing: Image TransformationChapter 4 Image Processing: Image Transformation
Chapter 4 Image Processing: Image Transformation
 
Convolutional Neural Network Models - Deep Learning
Convolutional Neural Network Models - Deep LearningConvolutional Neural Network Models - Deep Learning
Convolutional Neural Network Models - Deep Learning
 
gaussian filter seminar ppt
gaussian filter seminar pptgaussian filter seminar ppt
gaussian filter seminar ppt
 
Understanding neural radiance fields
Understanding neural radiance fieldsUnderstanding neural radiance fields
Understanding neural radiance fields
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
6.frequency domain image_processing
6.frequency domain image_processing6.frequency domain image_processing
6.frequency domain image_processing
 
Object Detection using Deep Neural Networks
Object Detection using Deep Neural NetworksObject Detection using Deep Neural Networks
Object Detection using Deep Neural Networks
 
Wavelet based image compression technique
Wavelet based image compression techniqueWavelet based image compression technique
Wavelet based image compression technique
 
Computer vision - images and image filtering
Computer vision - images and image filtering Computer vision - images and image filtering
Computer vision - images and image filtering
 
Convolutional Neural Network
Convolutional Neural NetworkConvolutional Neural Network
Convolutional Neural Network
 
CNN and its applications by ketaki
CNN and its applications by ketakiCNN and its applications by ketaki
CNN and its applications by ketaki
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)
The Rendering Technology of 'Lords of the Fallen' (Game Connection Europe 2014)
 
Enhancement in frequency domain
Enhancement in frequency domainEnhancement in frequency domain
Enhancement in frequency domain
 
Digital image processing img smoothning
Digital image processing img smoothningDigital image processing img smoothning
Digital image processing img smoothning
 
Image processing, Noise, Noise Removal filters
Image processing, Noise, Noise Removal filtersImage processing, Noise, Noise Removal filters
Image processing, Noise, Noise Removal filters
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
Digital Image Processing: Digital Image Fundamentals
Digital Image Processing: Digital Image FundamentalsDigital Image Processing: Digital Image Fundamentals
Digital Image Processing: Digital Image Fundamentals
 
Object detection - RCNNs vs Retinanet
Object detection - RCNNs vs RetinanetObject detection - RCNNs vs Retinanet
Object detection - RCNNs vs Retinanet
 

Similar to Final presentation optical flow estimation with DL

DALL-E.pdf
DALL-E.pdfDALL-E.pdf
DALL-E.pdfdsfajkh
 
IIBMP2019 講演資料「オープンソースで始める深層学習」
IIBMP2019 講演資料「オープンソースで始める深層学習」IIBMP2019 講演資料「オープンソースで始める深層学習」
IIBMP2019 講演資料「オープンソースで始める深層学習」Preferred Networks
 
VRP2013 - Comp Aspects VRP
VRP2013 - Comp Aspects VRPVRP2013 - Comp Aspects VRP
VRP2013 - Comp Aspects VRPVictor Pillac
 
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...IOSR Journals
 
Landmark Retrieval & Recognition
Landmark Retrieval & RecognitionLandmark Retrieval & Recognition
Landmark Retrieval & Recognitionkenluck2001
 
rit seminars-privacy assured outsourcing of image reconstruction services in ...
rit seminars-privacy assured outsourcing of image reconstruction services in ...rit seminars-privacy assured outsourcing of image reconstruction services in ...
rit seminars-privacy assured outsourcing of image reconstruction services in ...thahirakabeer
 
Online advertising and large scale model fitting
Online advertising and large scale model fittingOnline advertising and large scale model fitting
Online advertising and large scale model fittingWush Wu
 
Fast Fingerprint Classification with Deep Neural Networks
Fast Fingerprint Classification with Deep Neural NetworksFast Fingerprint Classification with Deep Neural Networks
Fast Fingerprint Classification with Deep Neural NetworksDaniel Michelsanti
 
Semantic Segmentation on Satellite Imagery
Semantic Segmentation on Satellite ImagerySemantic Segmentation on Satellite Imagery
Semantic Segmentation on Satellite ImageryRAHUL BHOJWANI
 
B Eng Final Year Project Presentation
B Eng Final Year Project PresentationB Eng Final Year Project Presentation
B Eng Final Year Project Presentationjesujoseph
 
Toronto meetup 20190917
Toronto meetup 20190917Toronto meetup 20190917
Toronto meetup 20190917Bill Liu
 
One-Pass Clustering Superpixels
One-Pass Clustering SuperpixelsOne-Pass Clustering Superpixels
One-Pass Clustering SuperpixelsKesavan Yogarajah
 
AI optimizing HPC simulations (presentation from 6th EULAG Workshop)
AI optimizing HPC simulations (presentation from  6th EULAG Workshop)AI optimizing HPC simulations (presentation from  6th EULAG Workshop)
AI optimizing HPC simulations (presentation from 6th EULAG Workshop)byteLAKE
 
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759Vandna Sambyal
 
Runtime Performance Optimizations for an OpenFOAM Simulation
Runtime Performance Optimizations for an OpenFOAM SimulationRuntime Performance Optimizations for an OpenFOAM Simulation
Runtime Performance Optimizations for an OpenFOAM SimulationFisnik Kraja
 
1c03projectlinkedin
1c03projectlinkedin1c03projectlinkedin
1c03projectlinkedinKeyur Patel
 
The Principle Of Ultrasound Imaging System
The Principle Of Ultrasound Imaging SystemThe Principle Of Ultrasound Imaging System
The Principle Of Ultrasound Imaging SystemMelissa Luster
 
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...Mohamed Elawady
 

Similar to Final presentation optical flow estimation with DL (20)

DALL-E.pdf
DALL-E.pdfDALL-E.pdf
DALL-E.pdf
 
IIBMP2019 講演資料「オープンソースで始める深層学習」
IIBMP2019 講演資料「オープンソースで始める深層学習」IIBMP2019 講演資料「オープンソースで始める深層学習」
IIBMP2019 講演資料「オープンソースで始める深層学習」
 
VRP2013 - Comp Aspects VRP
VRP2013 - Comp Aspects VRPVRP2013 - Comp Aspects VRP
VRP2013 - Comp Aspects VRP
 
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...
High Speed and Time Efficient 1-D DWT on Xilinx Virtex4 DWT Using 9/7 Filter ...
 
Landmark Retrieval & Recognition
Landmark Retrieval & RecognitionLandmark Retrieval & Recognition
Landmark Retrieval & Recognition
 
rit seminars-privacy assured outsourcing of image reconstruction services in ...
rit seminars-privacy assured outsourcing of image reconstruction services in ...rit seminars-privacy assured outsourcing of image reconstruction services in ...
rit seminars-privacy assured outsourcing of image reconstruction services in ...
 
Online advertising and large scale model fitting
Online advertising and large scale model fittingOnline advertising and large scale model fitting
Online advertising and large scale model fitting
 
Fast Fingerprint Classification with Deep Neural Networks
Fast Fingerprint Classification with Deep Neural NetworksFast Fingerprint Classification with Deep Neural Networks
Fast Fingerprint Classification with Deep Neural Networks
 
Semantic Segmentation on Satellite Imagery
Semantic Segmentation on Satellite ImagerySemantic Segmentation on Satellite Imagery
Semantic Segmentation on Satellite Imagery
 
B Eng Final Year Project Presentation
B Eng Final Year Project PresentationB Eng Final Year Project Presentation
B Eng Final Year Project Presentation
 
Electrolux meetup
Electrolux meetupElectrolux meetup
Electrolux meetup
 
Toronto meetup 20190917
Toronto meetup 20190917Toronto meetup 20190917
Toronto meetup 20190917
 
One-Pass Clustering Superpixels
One-Pass Clustering SuperpixelsOne-Pass Clustering Superpixels
One-Pass Clustering Superpixels
 
AI optimizing HPC simulations (presentation from 6th EULAG Workshop)
AI optimizing HPC simulations (presentation from  6th EULAG Workshop)AI optimizing HPC simulations (presentation from  6th EULAG Workshop)
AI optimizing HPC simulations (presentation from 6th EULAG Workshop)
 
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759
Optimizedfeedforwardnetworkofcnnwithxnorv5 180321130759
 
Optimized feedforward network of cnn with xnor v5
Optimized feedforward network of cnn with xnor v5Optimized feedforward network of cnn with xnor v5
Optimized feedforward network of cnn with xnor v5
 
Runtime Performance Optimizations for an OpenFOAM Simulation
Runtime Performance Optimizations for an OpenFOAM SimulationRuntime Performance Optimizations for an OpenFOAM Simulation
Runtime Performance Optimizations for an OpenFOAM Simulation
 
1c03projectlinkedin
1c03projectlinkedin1c03projectlinkedin
1c03projectlinkedin
 
The Principle Of Ultrasound Imaging System
The Principle Of Ultrasound Imaging SystemThe Principle Of Ultrasound Imaging System
The Principle Of Ultrasound Imaging System
 
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...
(Msc Thesis) Sparse Coral Classification Using Deep Convolutional Neural Netw...
 

More from LeapMind Inc

[Icml2019] mix hop higher-order graph convolutional architectures via spars...
[Icml2019]  mix hop  higher-order graph convolutional architectures via spars...[Icml2019]  mix hop  higher-order graph convolutional architectures via spars...
[Icml2019] mix hop higher-order graph convolutional architectures via spars...LeapMind Inc
 
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...LeapMind Inc
 
[Icml2019] parameter efficient training of deep convolutional neural network...
[Icml2019] parameter efficient training of  deep convolutional neural network...[Icml2019] parameter efficient training of  deep convolutional neural network...
[Icml2019] parameter efficient training of deep convolutional neural network...LeapMind Inc
 
エッジ向けDeepLearningプロジェクトで必要なこと
エッジ向けDeepLearningプロジェクトで必要なことエッジ向けDeepLearningプロジェクトで必要なこと
エッジ向けDeepLearningプロジェクトで必要なことLeapMind Inc
 
20190227[EDLS]JAL's INNOVATION エアラインのAI活用
20190227[EDLS]JAL's INNOVATION エアラインのAI活用20190227[EDLS]JAL's INNOVATION エアラインのAI活用
20190227[EDLS]JAL's INNOVATION エアラインのAI活用LeapMind Inc
 
E20190227[EDLS]インテル®︎FPGAによるエッジAI
E20190227[EDLS]インテル®︎FPGAによるエッジAIE20190227[EDLS]インテル®︎FPGAによるエッジAI
E20190227[EDLS]インテル®︎FPGAによるエッジAILeapMind Inc
 
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係LeapMind Inc
 
20180831 [DeLTA TECH] 深く青い脂
20180831 [DeLTA TECH] 深く青い脂20180831 [DeLTA TECH] 深く青い脂
20180831 [DeLTA TECH] 深く青い脂LeapMind Inc
 
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜LeapMind Inc
 
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)LeapMind Inc
 
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.120180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1LeapMind Inc
 
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説LeapMind Inc
 
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGA
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGAAn Introduction of DNN Compression Technology and Hardware Acceleration on FPGA
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGALeapMind Inc
 
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会LeapMind Inc
 
JUIZ DLK 組込み向けDeep Learningコンパイラ
JUIZ DLK 組込み向けDeep LearningコンパイラJUIZ DLK 組込み向けDeep Learningコンパイラ
JUIZ DLK 組込み向けDeep LearningコンパイラLeapMind Inc
 

More from LeapMind Inc (16)

[Icml2019] mix hop higher-order graph convolutional architectures via spars...
[Icml2019]  mix hop  higher-order graph convolutional architectures via spars...[Icml2019]  mix hop  higher-order graph convolutional architectures via spars...
[Icml2019] mix hop higher-order graph convolutional architectures via spars...
 
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...
[Icml2019]LIT: Learned Intermediate Representation Training for Model Compres...
 
[Icml2019] parameter efficient training of deep convolutional neural network...
[Icml2019] parameter efficient training of  deep convolutional neural network...[Icml2019] parameter efficient training of  deep convolutional neural network...
[Icml2019] parameter efficient training of deep convolutional neural network...
 
エッジ向けDeepLearningプロジェクトで必要なこと
エッジ向けDeepLearningプロジェクトで必要なことエッジ向けDeepLearningプロジェクトで必要なこと
エッジ向けDeepLearningプロジェクトで必要なこと
 
20190227[EDLS]JAL's INNOVATION エアラインのAI活用
20190227[EDLS]JAL's INNOVATION エアラインのAI活用20190227[EDLS]JAL's INNOVATION エアラインのAI活用
20190227[EDLS]JAL's INNOVATION エアラインのAI活用
 
E20190227[EDLS]インテル®︎FPGAによるエッジAI
E20190227[EDLS]インテル®︎FPGAによるエッジAIE20190227[EDLS]インテル®︎FPGAによるエッジAI
E20190227[EDLS]インテル®︎FPGAによるエッジAI
 
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係
20190227[EDLS]進化するAI on Edge 〜 CloudとEdgeの最適な関係
 
20180831 [DeLTA TECH] 深く青い脂
20180831 [DeLTA TECH] 深く青い脂20180831 [DeLTA TECH] 深く青い脂
20180831 [DeLTA TECH] 深く青い脂
 
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜
20180831 [DeLTA TECH] 新・深層の世紀 〜第3集 ディープラーニング・時代はAIを求めた 〜
 
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)
20180831 [DeLTA TECH] DeLTA-Liteを支える技術(システム構成編)
 
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.120180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1
20180831 [DeLTA TECH] DeLTA-FamilyによるIndustry4.1
 
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説
20180613 [TensorFlow分散学習] Horovodによる分散学習の実装方法と解説
 
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGA
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGAAn Introduction of DNN Compression Technology and Hardware Acceleration on FPGA
An Introduction of DNN Compression Technology and Hardware Acceleration on FPGA
 
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会
2018年1月19日開催 IoTビジネス共創ラボ 第6回勉強会
 
Pitch v2.2
Pitch v2.2Pitch v2.2
Pitch v2.2
 
JUIZ DLK 組込み向けDeep Learningコンパイラ
JUIZ DLK 組込み向けDeep LearningコンパイラJUIZ DLK 組込み向けDeep Learningコンパイラ
JUIZ DLK 組込み向けDeep Learningコンパイラ
 

Recently uploaded

Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSRajkumarAkumalla
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxupamatechverse
 
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...Soham Mondal
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...ranjana rawat
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Dr.Costas Sachpazis
 
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...srsj9000
 
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escortsranjana rawat
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxJoão Esperancinha
 
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escortsranjana rawat
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)Suman Mia
 
chaitra-1.pptx fake news detection using machine learning
chaitra-1.pptx  fake news detection using machine learningchaitra-1.pptx  fake news detection using machine learning
chaitra-1.pptx fake news detection using machine learningmisbanausheenparvam
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxpranjaldaimarysona
 
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130Suhani Kapoor
 
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝soniya singh
 
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSAPPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSKurinjimalarL3
 
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxthe ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxhumanexperienceaaa
 

Recently uploaded (20)

Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptx
 
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
 
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Meera Call 7001035870 Meet With Nagpur Escorts
 
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
 
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur EscortsHigh Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
High Profile Call Girls Nagpur Isha Call 7001035870 Meet With Nagpur Escorts
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
 
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
(MEERA) Dapodi Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Escorts
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
 
chaitra-1.pptx fake news detection using machine learning
chaitra-1.pptx  fake news detection using machine learningchaitra-1.pptx  fake news detection using machine learning
chaitra-1.pptx fake news detection using machine learning
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptx
 
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
VIP Call Girls Service Kondapur Hyderabad Call +91-8250192130
 
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
 
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
 
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSAPPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
 
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCRCall Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
 
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxthe ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
 

Final presentation optical flow estimation with DL

  • 1. Optical flow estimation with Blueoil K-Inoue @ki42 & Oscar @wang Blueoil Division 1
  • 2. What is optical flow estimation? ● Optical flow... represents the apparent motion of objects. ● Optical flow estimation... can predict the movement of objects in a video. Miloud, Hadj achour. (2017). Fragmentation de métal liquide dans l’eau. https://www.codeproject.com/Articles/1192205/Capturing-motion-from-video-using-the-Emgu-CV-libr 2
  • 3. Optical flow estimation is important ● Widely used by insects and birds ● Practical usage ○ Analyze motion ○ Avoid collision ○ Assist in navigation ● Real-world Applications ○ Video/Motion classification ○ Navigation assistance ■ Self driving cars ■ Drones https://nanonets.com/blog/optical-flow/ 3
  • 4. DL approaches are increasing 4 ● 2015. FlowNet S (Simple) ● 2015. FlowNet C (Correlated) ● 2016. FlowNet 2 ● 2018. LiteFlowNet ● ... Color coding https://www.youtube.com/watch?v=k_wkDLJ8lJE https://www.youtube.com/watch?v=pfQ0zFwv-hM
  • 5. However... ● Existing DL approaches require GPU to execute 👎 High power consumption 👎 Low runtime speed on CPU environment ● We propose LmFlowNet S 👍 ○ Modification of FlowNet S [P. Fischer+ 2015] ○ Goals: ■ Edge Computing ■ Run on FPGA-based accelerator ■ Use quantization to reduce inference time while achieving good prediction performance 5
  • 6. FlowNet S vs LmFlowNet S 6
  • 7. Network of FlowNet S 7 [N, 384, 512, 6] [N, 384, 512, 2] Detailed ops inside each color block is shown in appendix
  • 8. Network of FlowNet S 8 [N, 384, 512, 6] [N, 384, 512, 2] Encoder Decoder Detailed ops inside each color block is shown in appendix
  • 9. Network of FlowNet S 9 [N, 384, 512, 6] [N, 384, 512, 2] Detailed ops inside each color block is shown in appendix Not supported by Blueoil DLK
  • 10. 10 Network of LmFlowNet S (DLK supported) [N, 384, 512, 6] [N, 384, 512, 2] Detailed ops inside each color block is shown in appendix
  • 11. 11 Network of LmFlowNet S [N, 384, 512, 6] [N, 384, 512, 2] Quantized Detailed quantization inside each color block is shown in appendix
  • 12. Loss function: End Point Error (EPE) 12 (x1 ,y1 ) (x2 ,y2 ) EPEflow2 EPEflow3 EPEflow4 EPEflow5 EPEflow6 Weighted EPE = 0.32 * EPEflow2 + 0.08 * EPEflow3 + 0.02 * EPEflow4 + 0.01 * EPEflow5 + 0.005 * EPEflow6 Down- sampled Ground Truth Training hyper-parameters are shown in appendix 24 x 32 48 x 6496 x 128 192 x 256 12 x 16
  • 13. Artificial dataset: Flying Chairs 13 ● Dataset Name Frame pairs Train validation ratio size Flying Chairs 22,872 9:1 30GB ● Data Augmentation ○ Crop, Rotate, Translate, FlipLeftRight, FlipTopBottom ○ Gaussian noise, Brightness, Contrast, Gamma, and Color Parameters used in data augmentation are shown in appendix https://arxiv.org/pdf/1504.06852.pdf
  • 15. 15 Results - Avg. EPE & Inference time Method Avg. EPE (pixel) (Flying Chairs) Inference time per frame (ms) [1] CPU (dlk-convert) GPU (tensorflow) FlowNet S 2.94 - 11.65 LmFlowNet S 5.33 1360.49 13.81 LmFlowNet S Quantized 9.01 637.467 17.60 [1] CPU and GPU specs available in appedix FlowNet S Trained for 1.2M LmFlowNet S Trained for 400K LmFlowNet S Quantized Trained for 400K
  • 17. 17 Live demonstration ● Three demonstration ○ FlowNet S ○ LmFlowNet S ○ LmFlowNet S Quantized ● NOTE: Running on GPU (not on CPU / FPGA) ○ Failed to run on CPU/FPGA due to several problems 😢 ■ etc. segmentation fault, memory error... ○ Fixing and debugging them in the future 👊
  • 18. 18
  • 20. Challenges ● Training takes a very long time ( > 2 weeks...😢) ○ Heavy data augmentation & pre-processing ■ Pre-processing on GPU is not supported now. ● Unique network structure, not compatible with Blueoil ○ Input is a stack of 2 images (6 channels) ○ Multiple and branched outputs ● DLK Limitation. No documentation. 🤯 ○ No support for kernel size 7x7, 5x5 ○ No support for Conv2dTranspose ○ Cannot concat quantized value and float together ○ Requires the depth of Space2Depth to be 32 * N 20
  • 21. Thank you for your attention! our source code: https://github.com/ki-lm/blueoil/tree/lmflownets 21
  • 22. 22
  • 24. 24
  • 25. LmFlowNet S | Training & Hyper-params ● Optimizer: Adam ● Max steps: 1200k ● Fixed parameters in Adam: β1 =0.9 and β2 = 0.999. ● Batch size: 8 ● Learning rate ○ values: [1e-4 , 5e-5 , 2.5e-5 ,1.25e-5 ,6.25e-6 ] ○ boundaries: [400K, 600K, 800K, 1000K] ● Learning rate for quantization ○ values: [1.25e-5 , 1e-4 , 5e-5 , 2.5e-5 ,1.25e-5 ,6.25e-6 ] ○ boundaries: [50K, 400K, 600K, 800K, 1000K] 25 Tokunaga Scheduling 😆
  • 26. LmFlowNet S | Data Augmentation ● Translation: [20%, 20%] of the image width for x and y ● Rotation: [17o , 17o ] ● Scaling: [0.9, 2.0] ● Gaussian noise: sigma uniformly sampled from [0, 0.04] ● Contrast: [0.8, 0.4] ● Multiplicative color changes to the RGB channels per image: [0.5, 2] ● Gamma values: [0.7, 1.5] ● Additive brightness changes: Gaussian with a sigma of 0.2 26
  • 27. 27 CPU & GPU specs ● CPU ○ Intel(R) Core(TM) i7-5500U CPU @ 2.40GHz ○ 4 cores, 8 threads ○ Inference time (ms) ■ LmFlowNet S: 1358.39, 1358.39, 1293.26, 1307.28, 1485.11 ■ LmFlowNet S Quant: 640.541, 640.732, 635.354, 635.354, 635.354 ● GPU ○ NVIDIA Tesla V100 on DGX-1 ○ Inference time (ms) ■ FlowNet S: 0.0111, 0.0105, 0.0116, 0.0117, 0.0119, 0.0118, 0.0121, 0.0118, 0.0116, 0.0124 ■ LmFlowNet S: 0.0137, 0.0140, 0.0137, 0.0136, 0.0141, 0.0135, 0.0139, 0.0138, 0.0136, 0.0142 ■ LmFlowNet S Quant: 0.0168, 0.0161, 0.0165, 0.0177, 0.0173, 0.0166, 0.0186, 0.0195, 0.0183, 0.0187
  • 28. Our FlowNet S versions 28 Version Architecture DLK support V1 (FlowNet S) Same as the paper X V2 7x7, 5x5 => 3x3 Striding 2 => SpaceToDepth X V3 Conv2dTranspose => ResizeNearestNeighbor + Conv2d ResizeBilinear => ResizeNearestNeighbor X V3 Quant. (LmFlowNet S) Quantize except first, last layer, and activation before last layer △ V4 Quant. Change all output depths from SpaceToDepth to 32 * N O
  • 29. List of source code links ● FlowNet S/C, and 2 (TensorFlow): https://github.com/sampepose/flownet2-tf/ ● FlowNet S/C (Original paper, Caffe): https://lmb.informatik.uni-freiburg.de/Publications/2015/DF IB15/ ● FlowNet 2 (Original paper, Caffe): https://github.com/lmb-freiburg/flownet2 29