SlideShare ist ein Scribd-Unternehmen logo
1 von 35
Downloaden Sie, um offline zu lesen
FlowNet: Learning Optical Flow
with Convolutional Networks
Hyeongmin Lee
Image and Video Pattern Recognition LAB
Electrical and Electronic Engineering Dept, Yonsei University
4th Semester
2019.12.15
What is Optical Flow?
What is Optical Flow?
◆ Optical Flow
연속한 두 Frame 사이에서 각 Pixel의 Motion을 나타내는 Vector Map (Pixel Displacement)
What is Optical Flow?
◆ Visualizing Optical Flow
<2-Dimensional Map>
• Color: Direction
• Saturation: Magnitude
Optical Flow Constraint
Optical Flow Constraint
◆ The Optical Flow Constraint
“Flow Vector로 이어진 두 Pixel 값은 동일하다.”
Optical Flow Constraint
◆ The Optical Flow Constraint
“Flow Vector로 이어진 두 Pixel 값은 동일하다.”
𝐼(𝑥, 𝑦, 𝑡) (𝑢, 𝑣)
Image Flow
Optical Flow Constraint
◆ The Optical Flow Constraint
“Flow Vector로 이어진 두 Pixel 값은 동일하다.”
𝐼(𝑥, 𝑦, 𝑡) (𝑢, 𝑣)
Image Flow
𝐼 𝑥, 𝑦, 𝑡 = 𝐼(𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1)
Optical Flow Constraint
◆ The Optical Flow Constraint
𝐼 𝑥, 𝑦, 𝑡 = 𝐼 𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1
= 𝐼 𝑥, 𝑦, 𝑡 + 1 +
𝜕𝐼
𝜕𝑥
𝑢 +
𝜕𝐼
𝜕𝑥
𝑣
𝐼 𝑥, 𝑦, 𝑡 + 1 − 𝐼 𝑥, 𝑦, 𝑡 +
𝜕𝐼
𝜕𝑥
𝑢 +
𝜕𝐼
𝜕𝑥
𝑣 = 0
𝐼𝑡 + 𝐼 𝑥 𝑢 + 𝐼 𝑦 𝑣 = 0
First-Order Taylor Approximation
The Optical Flow Constraint
Optical Flow Constraint
◆ Aperture Problem
𝐼𝑡 + 𝐼 𝑥 𝑢 + 𝐼 𝑦 𝑣 = 0
The Optical Flow Constraint
(Underdetermined)
We need some additional constraints!!
Optical Flow Estimation by Optimization
Optical Flow Estimation by Optimization
◆ Lucas-Kanade Method
𝐼𝑡
(1)
+ 𝐼 𝑥
(1)
𝑢 + 𝐼 𝑦
(1)
𝑣 = 0
𝐼𝑡
(2)
+ 𝐼 𝑥
(2)
𝑢 + 𝐼 𝑦
(2)
𝑣 = 0
𝐼𝑡
(3)
+ 𝐼 𝑥
(3)
𝑢 + 𝐼 𝑦
(3)
𝑣 = 0
𝐼𝑡
(𝑛)
+ 𝐼 𝑥
(𝑛)
𝑢 + 𝐼 𝑦
(𝑛)
𝑣 = 0
Optical Flow Estimation by Optimization
◆ Variational Method
Optical Flow ConstraintSmoothness Constraint
(Total Variation Loss)
Optical Flow Estimation by Optimization
◆ Variational Method – Quadratic Relaxation
For N Iterations
Optical Flow Estimation by Energy Minimization
◆ Large Displacement
𝐼 𝑥, 𝑦, 𝑡 = 𝐼 𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1
= 𝐼 𝑥, 𝑦, 𝑡 + 1 +
𝜕𝐼
𝜕𝑥
𝑢 +
𝜕𝐼
𝜕𝑥
𝑣 First-Order Taylor Approximation
Short Displacement가 전제됨!
➔ Large Displacement에 취약
✓ Coarse-to-Fine Method
Error Propagation
Optical Flow Estimation by Energy Minimization
◆ Large Displacement Optical Flow [TPAMI 2011]
Descriptor Loss
Descriptor Based Feature Matching을 Guide로 주어, Large Displacement에 대한 성능을 보완
Optical Flow Estimation by Energy Minimization
◆ Large Displacement Optical Flow [TPAMI 2011]
✓ Feature Matching using Descriptors
Large Displacement에 강인.
Sparse한 점에 대해서만 Matching 가능.
Optical Flow Estimation by Energy Minimization
◆ EpicFlow [CVPR 2015]
Coarse-to-Fine Interpolation 시에 Edge를 고려하여 Sharp한 결과를 얻음
Optical Flow Estimation by Energy Minimization
◆ EpicFlow [CVPR 2015]
Interpolation using Euclidean distance
Euclidean
edge
far
close
Interpolation using Geodesic distance
FlowNet [ICCV 2015]
FlowNet
◆ Why optimization based?
✓ Hardware와 Deep Learning의 기술적 Baseline 부족
✓ Hard to get ground truth (Lack of Dataset)
FlowNet
◆ Why optimization based?
✓ Hardware와 Deep Learning의 기술적 Baseline 부족
✓ Hard to get ground truth (Lack of Dataset)
GPU & Parallel Processing
Convolutional Neural Networks
만들자!
FlowNet
◆ Flying Chairs Dataset
Flicker DB + 3d Chairs
FlowNet
◆ FlowNet
FlowNet
◆ Correlation Layer
𝐱 𝟏
𝐱 𝟐
𝐷
𝐷
𝐾
𝐾
𝑊
𝐻
𝑊
𝐻
𝐷2
FlowNet
◆ Refinement Layer
EPE loss
EPE loss
EPE loss
EPE loss0.08
0.02
0.01
0.005
EPE(End Point Error) Loss: Estimated & GT간의 L2 Loss
FlowNet
◆ Results
FlowNet
◆ Results
FlowNet2.0 [CVPR 2017]
FlowNet2.0
◆ Change on Training Dataset & Scheduling
1. FlowNet에서 사용한 Flying Chair Dataset으로 pre-train
2. Mayer et al. 에서 제안한 Flying Things 3D Dataset으로 추가 학습
FlowNet2.0
◆ Change on Network Architecture
FlowNet2.0
◆ New Dataset & Architecture for Small Displacement
1. ChairsSDHom Dataset (like UCF101)
2. FlowNet-SD Block & Fusion Block
FlowNet2.0
◆ Results
FlowNet2.0
◆ Results
Thank You!

Weitere ähnliche Inhalte

Was ist angesagt?

ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]Dongmin Choi
 
PR-376: Softmax Splatting for Video Frame Interpolation
PR-376: Softmax Splatting for Video Frame InterpolationPR-376: Softmax Splatting for Video Frame Interpolation
PR-376: Softmax Splatting for Video Frame InterpolationHyeongmin Lee
 
Semantic Segmentation Methods using Deep Learning
Semantic Segmentation Methods using Deep LearningSemantic Segmentation Methods using Deep Learning
Semantic Segmentation Methods using Deep LearningSungjoon Choi
 
Emerging Properties in Self-Supervised Vision Transformers
Emerging Properties in Self-Supervised Vision TransformersEmerging Properties in Self-Supervised Vision Transformers
Emerging Properties in Self-Supervised Vision TransformersSungchul Kim
 
Survey on optical flow estimation with DL
Survey on optical flow estimation with DLSurvey on optical flow estimation with DL
Survey on optical flow estimation with DLLeapMind Inc
 
PR-231: A Simple Framework for Contrastive Learning of Visual Representations
PR-231: A Simple Framework for Contrastive Learning of Visual RepresentationsPR-231: A Simple Framework for Contrastive Learning of Visual Representations
PR-231: A Simple Framework for Contrastive Learning of Visual RepresentationsJinwon Lee
 
Explicit Density Models
Explicit Density ModelsExplicit Density Models
Explicit Density ModelsSangwoo Mo
 
[IBIS2017 講演] ディープラーニングによる画像変換
[IBIS2017 講演] ディープラーニングによる画像変換[IBIS2017 講演] ディープラーニングによる画像変換
[IBIS2017 講演] ディープラーニングによる画像変換Satoshi Iizuka
 
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolutioncvpaper. challenge
 
NetVLAD: CNN architecture for weakly supervised place recognition
NetVLAD:  CNN architecture for weakly supervised place recognitionNetVLAD:  CNN architecture for weakly supervised place recognition
NetVLAD: CNN architecture for weakly supervised place recognitionGeunhee Cho
 
Depth Fusion from RGB and Depth Sensors by Deep Learning
Depth Fusion from RGB and Depth Sensors by Deep LearningDepth Fusion from RGB and Depth Sensors by Deep Learning
Depth Fusion from RGB and Depth Sensors by Deep LearningYu Huang
 
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用SSII
 
Mask-RCNN for Instance Segmentation
Mask-RCNN for Instance SegmentationMask-RCNN for Instance Segmentation
Mask-RCNN for Instance SegmentationDat Nguyen
 
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)Masakazu Iwamura
 
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...Universitat Politècnica de Catalunya
 
FastDepth: Fast Monocular Depth Estimation on Embedded Systems
FastDepth: Fast Monocular Depth Estimation on Embedded SystemsFastDepth: Fast Monocular Depth Estimation on Embedded Systems
FastDepth: Fast Monocular Depth Estimation on Embedded Systemsharmonylab
 
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...Universitat Politècnica de Catalunya
 

Was ist angesagt? (20)

ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]ViT (Vision Transformer) Review [CDM]
ViT (Vision Transformer) Review [CDM]
 
PR-376: Softmax Splatting for Video Frame Interpolation
PR-376: Softmax Splatting for Video Frame InterpolationPR-376: Softmax Splatting for Video Frame Interpolation
PR-376: Softmax Splatting for Video Frame Interpolation
 
Semantic Segmentation Methods using Deep Learning
Semantic Segmentation Methods using Deep LearningSemantic Segmentation Methods using Deep Learning
Semantic Segmentation Methods using Deep Learning
 
Emerging Properties in Self-Supervised Vision Transformers
Emerging Properties in Self-Supervised Vision TransformersEmerging Properties in Self-Supervised Vision Transformers
Emerging Properties in Self-Supervised Vision Transformers
 
Survey on optical flow estimation with DL
Survey on optical flow estimation with DLSurvey on optical flow estimation with DL
Survey on optical flow estimation with DL
 
Deep Learning for Computer Vision: Object Detection (UPC 2016)
Deep Learning for Computer Vision: Object Detection (UPC 2016)Deep Learning for Computer Vision: Object Detection (UPC 2016)
Deep Learning for Computer Vision: Object Detection (UPC 2016)
 
PR-231: A Simple Framework for Contrastive Learning of Visual Representations
PR-231: A Simple Framework for Contrastive Learning of Visual RepresentationsPR-231: A Simple Framework for Contrastive Learning of Visual Representations
PR-231: A Simple Framework for Contrastive Learning of Visual Representations
 
Explicit Density Models
Explicit Density ModelsExplicit Density Models
Explicit Density Models
 
Mean Teacher
Mean TeacherMean Teacher
Mean Teacher
 
[IBIS2017 講演] ディープラーニングによる画像変換
[IBIS2017 講演] ディープラーニングによる画像変換[IBIS2017 講演] ディープラーニングによる画像変換
[IBIS2017 講演] ディープラーニングによる画像変換
 
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
 
NetVLAD: CNN architecture for weakly supervised place recognition
NetVLAD:  CNN architecture for weakly supervised place recognitionNetVLAD:  CNN architecture for weakly supervised place recognition
NetVLAD: CNN architecture for weakly supervised place recognition
 
Deep Learning for Computer Vision: Data Augmentation (UPC 2016)
Deep Learning for Computer Vision: Data Augmentation (UPC 2016)Deep Learning for Computer Vision: Data Augmentation (UPC 2016)
Deep Learning for Computer Vision: Data Augmentation (UPC 2016)
 
Depth Fusion from RGB and Depth Sensors by Deep Learning
Depth Fusion from RGB and Depth Sensors by Deep LearningDepth Fusion from RGB and Depth Sensors by Deep Learning
Depth Fusion from RGB and Depth Sensors by Deep Learning
 
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
SSII2021 [OS2-03] 自己教師あり学習における対照学習の基礎と応用
 
Mask-RCNN for Instance Segmentation
Mask-RCNN for Instance SegmentationMask-RCNN for Instance Segmentation
Mask-RCNN for Instance Segmentation
 
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)
ディープラーニングを用いた物体認識とその周辺 ~現状と課題~ (Revised on 18 July, 2018)
 
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...
Generative Adversarial Networks GAN - Xavier Giro - UPC TelecomBCN Barcelona ...
 
FastDepth: Fast Monocular Depth Estimation on Embedded Systems
FastDepth: Fast Monocular Depth Estimation on Embedded SystemsFastDepth: Fast Monocular Depth Estimation on Embedded Systems
FastDepth: Fast Monocular Depth Estimation on Embedded Systems
 
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...
Image classification on Imagenet (D1L4 2017 UPC Deep Learning for Computer Vi...
 

Ähnlich wie PR-214: FlowNet: Learning Optical Flow with Convolutional Networks

Optic flow estimation with deep learning
Optic flow estimation with deep learningOptic flow estimation with deep learning
Optic flow estimation with deep learningYu Huang
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingNAVER Engineering
 
2017 09-29 ndt loop closure
2017 09-29 ndt loop closure2017 09-29 ndt loop closure
2017 09-29 ndt loop closureiMorpheus ai
 
Acoustic echo cancellation
Acoustic echo cancellationAcoustic echo cancellation
Acoustic echo cancellationchintanajoshi
 
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)Tatsunori Taniai
 
The information and mutual information ration for counting image features and...
The information and mutual information ration for counting image features and...The information and mutual information ration for counting image features and...
The information and mutual information ration for counting image features and...alikhajegili
 
SVD and the Netflix Dataset
SVD and the Netflix DatasetSVD and the Netflix Dataset
SVD and the Netflix DatasetBen Mabey
 
Advanced cosine measures for collaborative filtering
Advanced cosine measures for collaborative filteringAdvanced cosine measures for collaborative filtering
Advanced cosine measures for collaborative filteringLoc Nguyen
 
Geometric wavelet transform for optical flow estimation algorithm
Geometric wavelet transform for optical flow estimation algorithmGeometric wavelet transform for optical flow estimation algorithm
Geometric wavelet transform for optical flow estimation algorithmijcga
 
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...Extend Your Journey: Introducing Signal Strength into Location-based Applicat...
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...Chih-Chuan Cheng
 
ct image quality
ct image qualityct image quality
ct image qualitydypradio
 
Presentation at SMI 2023
Presentation at SMI 2023Presentation at SMI 2023
Presentation at SMI 2023Joaquim Jorge
 
[PR12] Making Convolutional Networks Shift-Invariant Again
[PR12] Making Convolutional Networks Shift-Invariant Again[PR12] Making Convolutional Networks Shift-Invariant Again
[PR12] Making Convolutional Networks Shift-Invariant AgainHyeongmin Lee
 
Feasibility of EEG Super-Resolution Using Deep Convolutional Networks
Feasibility of EEG Super-Resolution Using Deep Convolutional NetworksFeasibility of EEG Super-Resolution Using Deep Convolutional Networks
Feasibility of EEG Super-Resolution Using Deep Convolutional NetworksSangjun Han
 
Arcis 2013 reservoir_bochure_
Arcis 2013 reservoir_bochure_Arcis 2013 reservoir_bochure_
Arcis 2013 reservoir_bochure_TGS
 
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...paperpublications3
 
Multiuser MIMO Vector Perturbation Precoding
Multiuser MIMO Vector Perturbation PrecodingMultiuser MIMO Vector Perturbation Precoding
Multiuser MIMO Vector Perturbation Precodingadeelrazi
 
Coherence enhancing shock filters
Coherence enhancing shock filtersCoherence enhancing shock filters
Coherence enhancing shock filtersYechan Ahn
 

Ähnlich wie PR-214: FlowNet: Learning Optical Flow with Convolutional Networks (20)

Optic flow estimation with deep learning
Optic flow estimation with deep learningOptic flow estimation with deep learning
Optic flow estimation with deep learning
 
Pr266
Pr266Pr266
Pr266
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch Matching
 
2017 09-29 ndt loop closure
2017 09-29 ndt loop closure2017 09-29 ndt loop closure
2017 09-29 ndt loop closure
 
Acoustic echo cancellation
Acoustic echo cancellationAcoustic echo cancellation
Acoustic echo cancellation
 
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)
Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017)
 
The information and mutual information ration for counting image features and...
The information and mutual information ration for counting image features and...The information and mutual information ration for counting image features and...
The information and mutual information ration for counting image features and...
 
SVD and the Netflix Dataset
SVD and the Netflix DatasetSVD and the Netflix Dataset
SVD and the Netflix Dataset
 
Advanced cosine measures for collaborative filtering
Advanced cosine measures for collaborative filteringAdvanced cosine measures for collaborative filtering
Advanced cosine measures for collaborative filtering
 
Geometric wavelet transform for optical flow estimation algorithm
Geometric wavelet transform for optical flow estimation algorithmGeometric wavelet transform for optical flow estimation algorithm
Geometric wavelet transform for optical flow estimation algorithm
 
CVPR presentation
CVPR presentationCVPR presentation
CVPR presentation
 
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...Extend Your Journey: Introducing Signal Strength into Location-based Applicat...
Extend Your Journey: Introducing Signal Strength into Location-based Applicat...
 
ct image quality
ct image qualityct image quality
ct image quality
 
Presentation at SMI 2023
Presentation at SMI 2023Presentation at SMI 2023
Presentation at SMI 2023
 
[PR12] Making Convolutional Networks Shift-Invariant Again
[PR12] Making Convolutional Networks Shift-Invariant Again[PR12] Making Convolutional Networks Shift-Invariant Again
[PR12] Making Convolutional Networks Shift-Invariant Again
 
Feasibility of EEG Super-Resolution Using Deep Convolutional Networks
Feasibility of EEG Super-Resolution Using Deep Convolutional NetworksFeasibility of EEG Super-Resolution Using Deep Convolutional Networks
Feasibility of EEG Super-Resolution Using Deep Convolutional Networks
 
Arcis 2013 reservoir_bochure_
Arcis 2013 reservoir_bochure_Arcis 2013 reservoir_bochure_
Arcis 2013 reservoir_bochure_
 
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...
Hybrid Quantum Genetic Particle Swarm Optimization Algorithm For Solving Opti...
 
Multiuser MIMO Vector Perturbation Precoding
Multiuser MIMO Vector Perturbation PrecodingMultiuser MIMO Vector Perturbation Precoding
Multiuser MIMO Vector Perturbation Precoding
 
Coherence enhancing shock filters
Coherence enhancing shock filtersCoherence enhancing shock filters
Coherence enhancing shock filters
 

Mehr von Hyeongmin Lee

PR-455: CoTracker: It is Better to Track Together
PR-455: CoTracker: It is Better to Track TogetherPR-455: CoTracker: It is Better to Track Together
PR-455: CoTracker: It is Better to Track TogetherHyeongmin Lee
 
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...Hyeongmin Lee
 
PR-420: Scalable Model Compression by Entropy Penalized Reparameterization
PR-420: Scalable Model Compression by Entropy Penalized ReparameterizationPR-420: Scalable Model Compression by Entropy Penalized Reparameterization
PR-420: Scalable Model Compression by Entropy Penalized ReparameterizationHyeongmin Lee
 
PR-409: Denoising Diffusion Probabilistic Models
PR-409: Denoising Diffusion Probabilistic ModelsPR-409: Denoising Diffusion Probabilistic Models
PR-409: Denoising Diffusion Probabilistic ModelsHyeongmin Lee
 
PR-395: Variational Image Compression with a Scale Hyperprior
PR-395: Variational Image Compression with a Scale HyperpriorPR-395: Variational Image Compression with a Scale Hyperprior
PR-395: Variational Image Compression with a Scale HyperpriorHyeongmin Lee
 
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...Hyeongmin Lee
 
PR-365: Fast object detection in compressed video
PR-365: Fast object detection in compressed videoPR-365: Fast object detection in compressed video
PR-365: Fast object detection in compressed videoHyeongmin Lee
 
PR-340: DVC: An End-to-end Deep Video Compression Framework
PR-340: DVC: An End-to-end Deep Video Compression FrameworkPR-340: DVC: An End-to-end Deep Video Compression Framework
PR-340: DVC: An End-to-end Deep Video Compression FrameworkHyeongmin Lee
 
PR-328: End-to-End Optimized Image Compression
PR-328: End-to-End OptimizedImage CompressionPR-328: End-to-End OptimizedImage Compression
PR-328: End-to-End Optimized Image CompressionHyeongmin Lee
 
PR-315: Taming Transformers for High-Resolution Image Synthesis
PR-315: Taming Transformers for High-Resolution Image SynthesisPR-315: Taming Transformers for High-Resolution Image Synthesis
PR-315: Taming Transformers for High-Resolution Image SynthesisHyeongmin Lee
 
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View SynthesisPR-302: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View SynthesisHyeongmin Lee
 
PR-252: Making Convolutional Networks Shift-Invariant Again
PR-252: Making Convolutional Networks Shift-Invariant AgainPR-252: Making Convolutional Networks Shift-Invariant Again
PR-252: Making Convolutional Networks Shift-Invariant AgainHyeongmin Lee
 
PR-240: Modulating Image Restoration with Continual Levels via Adaptive Featu...
PR-240: Modulating Image Restoration with Continual Levels viaAdaptive Featu...PR-240: Modulating Image Restoration with Continual Levels viaAdaptive Featu...
PR-240: Modulating Image Restoration with Continual Levels via Adaptive Featu...Hyeongmin Lee
 
Latest Frame interpolation Algorithms
Latest Frame interpolation AlgorithmsLatest Frame interpolation Algorithms
Latest Frame interpolation AlgorithmsHyeongmin Lee
 
[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping
[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping
[Paper Review] Temporal Generative Adversarial Nets with Singular Value ClippingHyeongmin Lee
 
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...Hyeongmin Lee
 
[Paper Review] Video Frame Interpolation via Adaptive Convolution
[Paper Review] Video Frame Interpolation via Adaptive Convolution[Paper Review] Video Frame Interpolation via Adaptive Convolution
[Paper Review] Video Frame Interpolation via Adaptive ConvolutionHyeongmin Lee
 
[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients
[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients
[Paper Review] A spatio -Temporal Descriptor Based on 3D -GradientsHyeongmin Lee
 
[Paper Review] Unmasking the abnormal events in video
[Paper Review] Unmasking the abnormal events in video[Paper Review] Unmasking the abnormal events in video
[Paper Review] Unmasking the abnormal events in videoHyeongmin Lee
 
GAN with Mathematics
GAN with MathematicsGAN with Mathematics
GAN with MathematicsHyeongmin Lee
 

Mehr von Hyeongmin Lee (20)

PR-455: CoTracker: It is Better to Track Together
PR-455: CoTracker: It is Better to Track TogetherPR-455: CoTracker: It is Better to Track Together
PR-455: CoTracker: It is Better to Track Together
 
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...
PR-430: CLIP4Clip: An Empirical Study of CLIP for End to End Video Clip Retri...
 
PR-420: Scalable Model Compression by Entropy Penalized Reparameterization
PR-420: Scalable Model Compression by Entropy Penalized ReparameterizationPR-420: Scalable Model Compression by Entropy Penalized Reparameterization
PR-420: Scalable Model Compression by Entropy Penalized Reparameterization
 
PR-409: Denoising Diffusion Probabilistic Models
PR-409: Denoising Diffusion Probabilistic ModelsPR-409: Denoising Diffusion Probabilistic Models
PR-409: Denoising Diffusion Probabilistic Models
 
PR-395: Variational Image Compression with a Scale Hyperprior
PR-395: Variational Image Compression with a Scale HyperpriorPR-395: Variational Image Compression with a Scale Hyperprior
PR-395: Variational Image Compression with a Scale Hyperprior
 
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...
PR-386: Light Field Networks: Neural Scene Representations with Single-Evalua...
 
PR-365: Fast object detection in compressed video
PR-365: Fast object detection in compressed videoPR-365: Fast object detection in compressed video
PR-365: Fast object detection in compressed video
 
PR-340: DVC: An End-to-end Deep Video Compression Framework
PR-340: DVC: An End-to-end Deep Video Compression FrameworkPR-340: DVC: An End-to-end Deep Video Compression Framework
PR-340: DVC: An End-to-end Deep Video Compression Framework
 
PR-328: End-to-End Optimized Image Compression
PR-328: End-to-End OptimizedImage CompressionPR-328: End-to-End OptimizedImage Compression
PR-328: End-to-End Optimized Image Compression
 
PR-315: Taming Transformers for High-Resolution Image Synthesis
PR-315: Taming Transformers for High-Resolution Image SynthesisPR-315: Taming Transformers for High-Resolution Image Synthesis
PR-315: Taming Transformers for High-Resolution Image Synthesis
 
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View SynthesisPR-302: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
PR-302: NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
 
PR-252: Making Convolutional Networks Shift-Invariant Again
PR-252: Making Convolutional Networks Shift-Invariant AgainPR-252: Making Convolutional Networks Shift-Invariant Again
PR-252: Making Convolutional Networks Shift-Invariant Again
 
PR-240: Modulating Image Restoration with Continual Levels via Adaptive Featu...
PR-240: Modulating Image Restoration with Continual Levels viaAdaptive Featu...PR-240: Modulating Image Restoration with Continual Levels viaAdaptive Featu...
PR-240: Modulating Image Restoration with Continual Levels via Adaptive Featu...
 
Latest Frame interpolation Algorithms
Latest Frame interpolation AlgorithmsLatest Frame interpolation Algorithms
Latest Frame interpolation Algorithms
 
[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping
[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping
[Paper Review] Temporal Generative Adversarial Nets with Singular Value Clipping
 
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...
[Paper Review] A Middlebury Benchmark & Context-Aware Synthesis for Video Fra...
 
[Paper Review] Video Frame Interpolation via Adaptive Convolution
[Paper Review] Video Frame Interpolation via Adaptive Convolution[Paper Review] Video Frame Interpolation via Adaptive Convolution
[Paper Review] Video Frame Interpolation via Adaptive Convolution
 
[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients
[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients
[Paper Review] A spatio -Temporal Descriptor Based on 3D -Gradients
 
[Paper Review] Unmasking the abnormal events in video
[Paper Review] Unmasking the abnormal events in video[Paper Review] Unmasking the abnormal events in video
[Paper Review] Unmasking the abnormal events in video
 
GAN with Mathematics
GAN with MathematicsGAN with Mathematics
GAN with Mathematics
 

Kürzlich hochgeladen

Online electricity billing project report..pdf
Online electricity billing project report..pdfOnline electricity billing project report..pdf
Online electricity billing project report..pdfKamal Acharya
 
Engineering Drawing focus on projection of planes
Engineering Drawing focus on projection of planesEngineering Drawing focus on projection of planes
Engineering Drawing focus on projection of planesRAJNEESHKUMAR341697
 
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxS1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxSCMS School of Architecture
 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VDineshKumar4165
 
A Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityA Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityMorshed Ahmed Rahath
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.Kamal Acharya
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaOmar Fathy
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTbhaskargani46
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"mphochane1998
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptDineshKumar4165
 
DC MACHINE-Motoring and generation, Armature circuit equation
DC MACHINE-Motoring and generation, Armature circuit equationDC MACHINE-Motoring and generation, Armature circuit equation
DC MACHINE-Motoring and generation, Armature circuit equationBhangaleSonal
 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXssuser89054b
 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startQuintin Balsdon
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptNANDHAKUMARA10
 
kiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadkiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadhamedmustafa094
 
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKAR
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKARHAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKAR
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKARKOUSTAV SARKAR
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueBhangaleSonal
 
Work-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxWork-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxJuliansyahHarahap1
 

Kürzlich hochgeladen (20)

Online electricity billing project report..pdf
Online electricity billing project report..pdfOnline electricity billing project report..pdf
Online electricity billing project report..pdf
 
Engineering Drawing focus on projection of planes
Engineering Drawing focus on projection of planesEngineering Drawing focus on projection of planes
Engineering Drawing focus on projection of planes
 
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptxS1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
S1S2 B.Arch MGU - HOA1&2 Module 3 -Temple Architecture of Kerala.pptx
 
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced LoadsFEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - V
 
A Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityA Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna Municipality
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS Lambda
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPT
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
 
Thermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.pptThermal Engineering -unit - III & IV.ppt
Thermal Engineering -unit - III & IV.ppt
 
DC MACHINE-Motoring and generation, Armature circuit equation
DC MACHINE-Motoring and generation, Armature circuit equationDC MACHINE-Motoring and generation, Armature circuit equation
DC MACHINE-Motoring and generation, Armature circuit equation
 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
 
Design For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the startDesign For Accessibility: Getting it right from the start
Design For Accessibility: Getting it right from the start
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.ppt
 
kiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal loadkiln thermal load.pptx kiln tgermal load
kiln thermal load.pptx kiln tgermal load
 
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKAR
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKARHAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKAR
HAND TOOLS USED AT ELECTRONICS WORK PRESENTED BY KOUSTAV SARKAR
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
 
Call Girls in South Ex (delhi) call me [🔝9953056974🔝] escort service 24X7
Call Girls in South Ex (delhi) call me [🔝9953056974🔝] escort service 24X7Call Girls in South Ex (delhi) call me [🔝9953056974🔝] escort service 24X7
Call Girls in South Ex (delhi) call me [🔝9953056974🔝] escort service 24X7
 
Work-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptxWork-Permit-Receiver-in-Saudi-Aramco.pptx
Work-Permit-Receiver-in-Saudi-Aramco.pptx
 

PR-214: FlowNet: Learning Optical Flow with Convolutional Networks

  • 1. FlowNet: Learning Optical Flow with Convolutional Networks Hyeongmin Lee Image and Video Pattern Recognition LAB Electrical and Electronic Engineering Dept, Yonsei University 4th Semester 2019.12.15
  • 3. What is Optical Flow? ◆ Optical Flow 연속한 두 Frame 사이에서 각 Pixel의 Motion을 나타내는 Vector Map (Pixel Displacement)
  • 4. What is Optical Flow? ◆ Visualizing Optical Flow <2-Dimensional Map> • Color: Direction • Saturation: Magnitude
  • 6. Optical Flow Constraint ◆ The Optical Flow Constraint “Flow Vector로 이어진 두 Pixel 값은 동일하다.”
  • 7. Optical Flow Constraint ◆ The Optical Flow Constraint “Flow Vector로 이어진 두 Pixel 값은 동일하다.” 𝐼(𝑥, 𝑦, 𝑡) (𝑢, 𝑣) Image Flow
  • 8. Optical Flow Constraint ◆ The Optical Flow Constraint “Flow Vector로 이어진 두 Pixel 값은 동일하다.” 𝐼(𝑥, 𝑦, 𝑡) (𝑢, 𝑣) Image Flow 𝐼 𝑥, 𝑦, 𝑡 = 𝐼(𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1)
  • 9. Optical Flow Constraint ◆ The Optical Flow Constraint 𝐼 𝑥, 𝑦, 𝑡 = 𝐼 𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1 = 𝐼 𝑥, 𝑦, 𝑡 + 1 + 𝜕𝐼 𝜕𝑥 𝑢 + 𝜕𝐼 𝜕𝑥 𝑣 𝐼 𝑥, 𝑦, 𝑡 + 1 − 𝐼 𝑥, 𝑦, 𝑡 + 𝜕𝐼 𝜕𝑥 𝑢 + 𝜕𝐼 𝜕𝑥 𝑣 = 0 𝐼𝑡 + 𝐼 𝑥 𝑢 + 𝐼 𝑦 𝑣 = 0 First-Order Taylor Approximation The Optical Flow Constraint
  • 10. Optical Flow Constraint ◆ Aperture Problem 𝐼𝑡 + 𝐼 𝑥 𝑢 + 𝐼 𝑦 𝑣 = 0 The Optical Flow Constraint (Underdetermined) We need some additional constraints!!
  • 11. Optical Flow Estimation by Optimization
  • 12. Optical Flow Estimation by Optimization ◆ Lucas-Kanade Method 𝐼𝑡 (1) + 𝐼 𝑥 (1) 𝑢 + 𝐼 𝑦 (1) 𝑣 = 0 𝐼𝑡 (2) + 𝐼 𝑥 (2) 𝑢 + 𝐼 𝑦 (2) 𝑣 = 0 𝐼𝑡 (3) + 𝐼 𝑥 (3) 𝑢 + 𝐼 𝑦 (3) 𝑣 = 0 𝐼𝑡 (𝑛) + 𝐼 𝑥 (𝑛) 𝑢 + 𝐼 𝑦 (𝑛) 𝑣 = 0
  • 13. Optical Flow Estimation by Optimization ◆ Variational Method Optical Flow ConstraintSmoothness Constraint (Total Variation Loss)
  • 14. Optical Flow Estimation by Optimization ◆ Variational Method – Quadratic Relaxation For N Iterations
  • 15. Optical Flow Estimation by Energy Minimization ◆ Large Displacement 𝐼 𝑥, 𝑦, 𝑡 = 𝐼 𝑥 + 𝑢, 𝑦 + 𝑣, 𝑡 + 1 = 𝐼 𝑥, 𝑦, 𝑡 + 1 + 𝜕𝐼 𝜕𝑥 𝑢 + 𝜕𝐼 𝜕𝑥 𝑣 First-Order Taylor Approximation Short Displacement가 전제됨! ➔ Large Displacement에 취약 ✓ Coarse-to-Fine Method Error Propagation
  • 16. Optical Flow Estimation by Energy Minimization ◆ Large Displacement Optical Flow [TPAMI 2011] Descriptor Loss Descriptor Based Feature Matching을 Guide로 주어, Large Displacement에 대한 성능을 보완
  • 17. Optical Flow Estimation by Energy Minimization ◆ Large Displacement Optical Flow [TPAMI 2011] ✓ Feature Matching using Descriptors Large Displacement에 강인. Sparse한 점에 대해서만 Matching 가능.
  • 18. Optical Flow Estimation by Energy Minimization ◆ EpicFlow [CVPR 2015] Coarse-to-Fine Interpolation 시에 Edge를 고려하여 Sharp한 결과를 얻음
  • 19. Optical Flow Estimation by Energy Minimization ◆ EpicFlow [CVPR 2015] Interpolation using Euclidean distance Euclidean edge far close Interpolation using Geodesic distance
  • 21. FlowNet ◆ Why optimization based? ✓ Hardware와 Deep Learning의 기술적 Baseline 부족 ✓ Hard to get ground truth (Lack of Dataset)
  • 22. FlowNet ◆ Why optimization based? ✓ Hardware와 Deep Learning의 기술적 Baseline 부족 ✓ Hard to get ground truth (Lack of Dataset) GPU & Parallel Processing Convolutional Neural Networks 만들자!
  • 23. FlowNet ◆ Flying Chairs Dataset Flicker DB + 3d Chairs
  • 25. FlowNet ◆ Correlation Layer 𝐱 𝟏 𝐱 𝟐 𝐷 𝐷 𝐾 𝐾 𝑊 𝐻 𝑊 𝐻 𝐷2
  • 26. FlowNet ◆ Refinement Layer EPE loss EPE loss EPE loss EPE loss0.08 0.02 0.01 0.005 EPE(End Point Error) Loss: Estimated & GT간의 L2 Loss
  • 30. FlowNet2.0 ◆ Change on Training Dataset & Scheduling 1. FlowNet에서 사용한 Flying Chair Dataset으로 pre-train 2. Mayer et al. 에서 제안한 Flying Things 3D Dataset으로 추가 학습
  • 31. FlowNet2.0 ◆ Change on Network Architecture
  • 32. FlowNet2.0 ◆ New Dataset & Architecture for Small Displacement 1. ChairsSDHom Dataset (like UCF101) 2. FlowNet-SD Block & Fusion Block