deep learning for the high dynamic range imaging pipeline · deep learning for the high dynamic...
Post on 20-May-2020
22 Views
Preview:
TRANSCRIPT
Deep Learning for the High Dynamic
Range Imaging PipelineDemetris Marnerides
Warwick Centre of Predictive Modelling (WCPM)
The University of Warwick
D.Marnerides@warwick.ac.uk
Project Supervisors: Dr Kurt Debattista (Primary – WMG), Dr Igor Khovanov (Secondary – WCPM)
Contents
Introduction to HDR
Introduction to Deep Learning
Relevant Work
Motivation
ExpandNet
Results
Future Work
2WCPM Seminar Series, December 2017
Introduction to HDR
Low/Standard Dynamic Range (LDR)
▪ Limited Luminance range
▪ Limited Colour gamut
▪ 8 bit quantization [0-255]
High Dynamic Range (HDR)
▪ Real-World Lighting
▪ 32-bit floats
3WCPM Seminar Series, December 2017
Introduction to HDR (2)
Most content is LDR
HDR LDR straightforward (Tone Mapping)
Inverse is hard (LDR HDR)
▪ Expert knowledge / heuristics
▪ Quantization, clipping
▪ Non-linear local luminance shifts
Proposed data-driven solution
▪ Learn relevant information from data
4WCPM Seminar Series, December 2017
?
Artificial Neural Networks
Single hidden layer
𝒚 = 𝐶𝒇 𝐴𝒙+ 𝒃 + 𝒅
Activations: Sigmoid, Tanh, Rectifiers (ReLU, PReLU, ELU, SELU) ...
Find parameters that minimize some ‘loss’ between model and data
▪ Euclidean distance (least squares regression)
𝑖
ഥ𝒚𝑖 − 𝒚𝑖2
Stochastic Gradient Descent with Backpropagation
5WCPM Seminar Series, December 2017
Going Deeper
𝒚 = 𝐴𝑵+𝟏𝒇𝑵 𝐴𝑵 …𝒇𝟐 𝐴𝟐𝒇𝟏 𝐴𝟏𝒙+ 𝒃𝟏 + 𝒃𝟐 …+ 𝒃𝑵 + 𝒃𝑵+𝟏
Depth
▪ Exponentially more expressive with less parameters
▪ Computationally more efficient
▪ Aids generalization over memorization
6WCPM Seminar Series, December 2017
Convolutions
7WCPM Seminar Series, December 2017
𝑦1 = 𝑥1 𝑓1+ 𝑥2 𝑓2+ 𝑥3 𝑓3
Convolutions
8WCPM Seminar Series, December 2017
𝑦2 = 𝑥2 𝑓1+ 𝑥3 𝑓2+ 𝑥4 𝑓3
Convolutions
9WCPM Seminar Series, December 2017
𝑦3 = 𝑥3 𝑓1+ 𝑥4 𝑓2+ 𝑥5 𝑓3
Convolutions
10WCPM Seminar Series, December 2017
𝑦4 = 𝑥4 𝑓1+ 𝑥5 𝑓2+ 𝑥6 𝑓3
Convolutions
11WCPM Seminar Series, December 2017
Likewise for 2D vectors (matrices, images)
𝑓𝑖𝑙𝑡𝑒𝑟 =1 0 10 1 01 0 1
Classification
Modular improvements:
▪ e.g. residual connections
Auto-encoders
Deep Convolutional Neural Networks
12WCPM Seminar Series, December 2017
LeNet-5
Inverse problems in Imaging
13WCPM Seminar Series, December 2017
Colorful Image
ColorizationZhang et al., 2016
Globally and Locally
Consistent
Image CompletionIizuka et al., 2017
Motivation for a new architecture
14WCPM Seminar Series, December 2017
UNet-like architectures:
▪ Abstract representations
▪ Multiscale context
▪ However prone to artefacts
• E.g. from the pix2pix
Semantic Segmentation results
https://phillipi.github.io/pix2pix/images/cityscapes_cGAN_AtoB/latest_net_G_val/index.html
ExpandNet
15WCPM Seminar Series, December 2017
Workflow (training)
16WCPM Seminar Series, December 2017
Workflow (testing)
17WCPM Seminar Series, December 2017
1000 nits (cd/m2)
▪ PU-PSNR
▪ PU-SSIM
▪ PU-MSSSIM
▪ HDR-VDP
Significance
▪ p < 0.01
Results
18WCPM Seminar Series, December 2017
Results (2)
19WCPM Seminar Series, December 2017
HDR-VDP-2 – Detection Probability Maps
Results (3)
20WCPM Seminar Series, December 2017
Image comparisons with other CNNs
Branches
21WCPM Seminar Series, December 2017
All branches
Local (masked D + G) Dilated (masked L + G)
Branches (2)
22WCPM Seminar Series, December 2017
Training combinations of branches
Future Work
23WCPM Seminar Series, December 2017
Reducing compression artefacts
Hallucinate under/over exposed regions
▪ Generative Adversarial Networks (GANs)
HDR Super-resolution
LDR to HDR Video
▪ Recurrent Neural Networks
Thank you!
24WCPM Seminar Series, December 2017
PhD is funded by the EPSRC
top related