Image to Image Translation using GANs

Slides:



Advertisements
Similar presentations
Deep Learning and Neural Nets Spring 2015
Advertisements

Generalizing Backpropagation to Include Sparse Coding David M. Bradley and Drew Bagnell Robotics Institute Carnegie.
Unsupervised Learning With Neural Nets Deep Learning and Neural Nets Spring 2015.
Neural Network Introduction Hung-yi Lee. Review: Supervised Learning Training: Pick the “best” Function f * Training Data Model Testing: Hypothesis Function.
Tips for Training Neural Network
Neural Networks Vladimir Pleskonjić 3188/ /20 Vladimir Pleskonjić General Feedforward neural networks Inputs are numeric features Outputs are in.
Convolutional Restricted Boltzmann Machines for Feature Learning Mohammad Norouzi Advisor: Dr. Greg Mori Simon Fraser University 27 Nov
Assignment 4: Deep Convolutional Neural Networks
Combining Models Foundations of Algorithms and Machine Learning (CS60020), IIT KGP, 2017: Indrajit Bhattacharya.
Machine Learning Supervised Learning Classification and Regression
Conditional Generative Adversarial Networks
Generative Adversarial Nets
Unsupervised Learning of Video Representations using LSTMs
Generative Adversarial Network (GAN)
Environment Generation with GANs
Deep Neural Net Scenery Generation
Energy models and Deep Belief Networks
Data Mining, Neural Network and Genetic Programming
Convolutional Neural Fabrics by Shreyas Saxena, Jakob Verbeek
Applications of Deep Learning and how to get started with implementation of deep learning Presentation By : Manaswi Advisor : Dr.Chinmay.
Matt Gormley Lecture 16 October 24, 2016
Restricted Boltzmann Machines for Classification
Generative Adversarial Networks
CSCI 5922 Neural Networks and Deep Learning Generative Adversarial Networks Mike Mozer Department of Computer Science and Institute of Cognitive Science.
Inception and Residual Architecture in Deep Convolutional Networks
Classification with Perceptrons Reading:
Chaoyun Zhang, Xi Ouyang, and Paul Patras
Synthesis of X-ray Projections via Deep Learning
CNN Demo LIU Pengpeng.
Classification of Hand-Written Digits Using Scattering Convolutional Network Dongmian Zou Advisor: Professor Radu Balan.
ECE 6504 Deep Learning for Perception
Deep learning and applications to Natural language processing
Generative adversarial networks (GANs) for edge detection
Authors: Jun-Yan Zhu*, Taesun Park*, Phillip Isola, Alexei A. Efros
FaceNet A Unified Embedding for Face Recognition and Clustering
Presenter: Hajar Emami
Adversarially Tuned Scene Generation
A brief introduction to neural network
Low Dose CT Image Denoising Using WGAN and Perceptual Loss
Proposed (MoDL-SToRM)
Normalized Cut Loss for Weakly-supervised CNN Segmentation
Face Recognition with Deep Learning Method
Paper Presentation Aryeh Zapinsky
PixelGAN Autoencoders
ALL YOU NEED IS A GOOD INIT
MXNet Internals Cyrus M. Vahid, Principal Solutions Architect,
network of simple neuron-like computing elements
Generative Adversarial Network
[Figure taken from googleblog
Image recognition: Defense adversarial attacks
GAN Applications.
Papers 15/08.
Ch4: Backpropagation (BP)
Lip movement Synthesis from Text
Zhedong Zheng, Liang Zheng and Yi Yang
实习生汇报 ——北邮 张安迪.
Textual Video Prediction
TPGAN overview.
Abnormally Detection
Sketch Object Prediction
Ch 14. Generative adversarial networks (GANs) for edge detection
Automatic Handwriting Generation
Introduction to Neural Networks
Deep Learning Libraries
Natalie Lang Tomer Malach
Ch4: Backpropagation (BP)
Angel A. Cantu, Nami Akazawa Department of Computer Science
End-to-End Speech-Driven Facial Animation with Temporal GANs
Generative adversarial networks (GANs)
An introduction to neural network and machine learning
Presentation transcript:

Image to Image Translation using GANs Shashank Verma

Executive Summary Objective: Domain Transfer between MNIST and SVHN datasets Overview: Generate fake MNIST images with random noise input. Generate SVHN-domain digits with MNIST and vice versa. Compare the different GAN architectures trained. MNIST (Handwritten Digits) SVHN (Street-View House Numbers)

Method – Part I First, we learn GANs (Generative Adversarial Networks) as a method to generate fake MNIST Images Vanilla GAN (proposed by Ian Goodfellow et al.) Deep Convolutional GAN (DCGAN)

Results (Part I) – Generated Images Vanilla GAN No. of Epochs: 200 Batch Size: 100 Optimizer: ADAM Loss: BCE Learning Rate = 0.0002 DCGAN No. of Epochs: 23 Batch Size: 100 Optimizer: ADAM Loss: BCE Learning Rate = 0.0002

Method – Part II MNIST -> SVHN (and vice versa) Domain Transfer Cycle Consistent Adversarial Nets (CycleGANs) - Discriminators output a probability Semi Supervised GAN (SGAN) – Discriminators output labels High Level CycleGAN Architecture

Results – Part II (CycleGAN) MNIST -> SVHN SVHN -> MNIST In most cases the correct digit information hasn’t been translated It is generating images that look like digits, and in the expected style Training Iterations: 91000 Loss: Reconstruction Loss + Squared Error Optimizer: ADAM Learning Rate: 0.0002 Momentum: 0.5

Results – Part II (SGAN) MNIST -> SVHN SVHN -> MNIST Much better results as compared to CycleGAN Most translated images have correct digit information Training Iterations: 100000 Loss: Cross Entropy Loss Optimizer: ADAM Learning Rate: 0.0002 Momentum: 0.5

Discussion DCGAN performed better than Vanilla GAN in generating fake MNIST images. DCGANs have more stable training dynamics as compared to Vanilla GANs. (Use Convolution layers where you can!) SGAN performed better than CycleGAN in the task of MNIST-SVHN domain transfer. In SGAN, the correct digit information was retained on more occasions as compared to CycleGAN. SGANs are a semi-supervised method requiring labeled data! GANs are hard to train! (especially Vanilla GAN) They suffer with “mode collapse”, “vanishing gradient” etc. A method to quantitatively compare domain transfer output is using off-the-shelf classifiers on the generated images.

Appendix Framework: PyTorch 0.4.1 Hardware: Euler Cluster (Intel® Haswell CPU (4 Cores in use) + Nvidia GeForce GTX-1080 (x 2)) Code Location: https://github.com/shashank3959/GAN