GitPlanet

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Created with love in Canada, visit hostnodejs.com today

Feel like to post an Ad? Learn Details
All Projects → vee-upatising → DLSS

vee-upatising / DLSS

Licence: other
Deep Learning Super Sampling with Deep Convolutional Generative Adversarial Networks.

Programming Languages

Jupyter Notebook
11667 projects
python
139335 projects - #7 most used programming language

Labels

deep-neural-networksdeep-learningkerasimage-processingcnngenerative-adversarial-networkgandcgankeras-tutorialsconvolutional-neural-networksganssuper-resolutionkeras-neural-networkscnn-kerasimage-renderingkeras-implementationskeras-gandlssdeep-learning-super-sampling

Projects that are alternatives of or similar to DLSS

Iseebetter
iSeeBetter: Spatio-Temporal Video Super Resolution using Recurrent-Generative Back-Projection Networks | Python3 | PyTorch | GANs | CNNs | ResNets | RNNs | Published in Springer Journal of Computational Visual Media, September 2020, Tsinghua University Press
Stars: ✭ 202 (+129.55%)
Mutual labels:  cnn, generative-adversarial-network, gan, gans, super-resolution
Fast Srgan
A Fast Deep Learning Model to Upsample Low Resolution Videos to High Resolution at 30fps
Stars: ✭ 417 (+373.86%)
Mutual labels:  cnn, generative-adversarial-network, gans, super-resolution
pytorch-gans
PyTorch implementation of GANs (Generative Adversarial Networks). DCGAN, Pix2Pix, CycleGAN, SRGAN
Stars: ✭ 21 (-76.14%)
Mutual labels:  generative-adversarial-network, dcgan, gans, super-resolution
Pytorch Srgan
A modern PyTorch implementation of SRGAN
Stars: ✭ 289 (+228.41%)
Mutual labels:  cnn, generative-adversarial-network, gan, super-resolution
Gans In Action
Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks
Stars: ✭ 748 (+750%)
Mutual labels:  generative-adversarial-network, gan, dcgan, gans
Generative adversarial networks 101
Keras implementations of Generative Adversarial Networks. GANs, DCGAN, CGAN, CCGAN, WGAN and LSGAN models with MNIST and CIFAR-10 datasets.
Stars: ✭ 138 (+56.82%)
Mutual labels:  generative-adversarial-network, gan, dcgan, gans
A Pytorch Tutorial To Super Resolution
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network | a PyTorch Tutorial to Super-Resolution
Stars: ✭ 157 (+78.41%)
Mutual labels:  generative-adversarial-network, gan, gans, super-resolution
Benchmarking Keras Pytorch
🔥 Reproducibly benchmarking Keras and PyTorch models
Stars: ✭ 346 (+293.18%)
Mutual labels:  cnn, keras-tutorials, cnn-keras
Tensorflow Tutorial
Tensorflow tutorial from basic to hard, 莫烦Python 中文AI教学
Stars: ✭ 4,122 (+4584.09%)
Mutual labels:  cnn, generative-adversarial-network, gan
GAN-Project-2018
GAN in Tensorflow to be run via Linux command line
Stars: ✭ 21 (-76.14%)
Mutual labels:  generative-adversarial-network, gan, dcgan
srgan
Pytorch implementation of "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"
Stars: ✭ 39 (-55.68%)
Mutual labels:  generative-adversarial-network, gan, super-resolution
Anogan Tf
Unofficial Tensorflow Implementation of AnoGAN (Anomaly GAN)
Stars: ✭ 218 (+147.73%)
Mutual labels:  generative-adversarial-network, gan, dcgan
Srgan Tensorflow
Tensorflow implementation of the SRGAN algorithm for single image super-resolution
Stars: ✭ 754 (+756.82%)
Mutual labels:  cnn, generative-adversarial-network, super-resolution
gans-collection.torch
Torch implementation of various types of GAN (e.g. DCGAN, ALI, Context-encoder, DiscoGAN, CycleGAN, EBGAN, LSGAN)
Stars: ✭ 53 (-39.77%)
Mutual labels:  generative-adversarial-network, dcgan, gans
Gif
GIF is a photorealistic generative face model with explicit 3D geometric and photometric control.
Stars: ✭ 233 (+164.77%)
Mutual labels:  generative-adversarial-network, gan, gans
Pytorch Cyclegan And Pix2pix
Image-to-Image Translation in PyTorch
Stars: ✭ 16,477 (+18623.86%)
Mutual labels:  generative-adversarial-network, gan, gans
Anime Face Gan Keras
A DCGAN to generate anime faces using custom mined dataset
Stars: ✭ 161 (+82.95%)
Mutual labels:  cnn, generative-adversarial-network, gan
Srgan
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
Stars: ✭ 2,641 (+2901.14%)
Mutual labels:  cnn, gan, super-resolution
coursera-gan-specialization
Programming assignments and quizzes from all courses within the GANs specialization offered by deeplearning.ai
Stars: ✭ 277 (+214.77%)
Mutual labels:  generative-adversarial-network, dcgan, gans
Image generator
DCGAN image generator 🖼️.
Stars: ✭ 173 (+96.59%)
Mutual labels:  generative-adversarial-network, gan, dcgan
View All Similar Projects ➔

Deep Learning Super Sampling

Using Deep Convolutional GANS to super sample images and increase their resolution.
pokemon1 pokemon2 pokemon3

How To Use This Repository

  • Tutorial

  • Requirements

    • Python 3
    • Keras (I use 2.3.1)
    • Tensorflow (I use 1.14.0)
    • Sklearn
    • Skimage
    • Numpy
    • Matplotlib
    • PIL

  • Documentation

    • DLSS GAN Training

      • This script is used to define the DCGAN class, train the Generative Adversarial Network, generate samples, and save the model at every epoch interval.

      • The Generator and Discriminator models were designed to be trained on an 8 GB GPU. If you have a less powerful GPU then decrease the conv_filter and kernel parameters accordingly.

      • User Specified Parameters:

        • input_path: File path pointing to the folder containing the low resolution dataset.
        • output_path: File path pointing to the folder containing the high resolution dataset.
        • input_dimensions: Dimensions of the images inside the low resolution dataset. The image sizes must be compatible meaning output_dimensions / input_dimensions is a multiple of 2.
        • output_dimensions: Dimensions of the images inside the high resolution dataset. The image sizes must be compatible meaning output_dimensions / input_dimensions is a multiple of 2.
        • super_sampling_ratio: Integer representing the ratio of the difference in size between the two image resolutions. This integer specifies how many times the Upsampling2D and MaxPooling2D layers are used in the models.
        • model_path: File path pointing to the folder where you want to save to model as well as generated samples.
        • interval: Integer representing how many epochs between saving your model.
        • epochs: Integer representing how many epochs to train the model.
        • batch: Integer representing how many images to train at one time.
        • conv_filters: Integer representing how many convolutional filters are used in each convolutional layer of the Generator and the Discriminator.
        • kernel: Tuple representing the size of the kernels used in the convolutional layers.
        • png: Boolean flag, set to True if the data has PNGs to remove alpha layer from images.

      • DCGAN Class:

        • __init__(self): The class is initialized by defining the dimensions of the input image as well as the output image. The Generator and Discriminator models get initialized using build_generator() and build_discriminator().
        • build_generator(self): Defines Generator model. The Convolutional and UpSampling2D layers increase the resolution of the image by a factor of super_sampling_ratio * 2. Gets called when the DCGAN class is initialized.
        • build_discriminator(self): Defines Discriminator model. The Convolutional and MaxPooling2D layers downsample from output_dimensions to 1 scalar prediction. Gets called when the DCGAN class is initialized.
        • load_data(self): Loads data from user specified file path, data_path. Reshapes images from input_path to have input_dimensions. Reshapes images from output_path to have output_dimensions. Gets called in the train() method.
        • train(self, epochs, batch_size, save_interval): Trains the Generative Adversarial Network. Each epoch trains the model using the entire dataset split up into chunks defined by batch_size. If epoch is at save_interval, then the method calls save_imgs() to generate samples and saves the model at the current epoch.
        • save_imgs(self, epoch, gen_imgs, interpolated): Saves the model and generates prediction samples for a given epoch at the user specified path, model_path. Each sample contains 8 interpolated images and Deep Learned Super Sampled images for comparison.

    • Load Model and Analyze Results

      • This script is used to super sample images using the Generator model trained by the DLSS GAN Training script.
      • The script will perform DLSS on all images inside the folder specified in dataset_path. You can insert frames of a video in here to create GIFs such as the one in the results section of this document.

      • User Specified Parameters:

        • input_dimensions: Dimensions of the image resolution the model takes as input.
        • output_dimensions: Dimensions of the image resolution the model takes as output.
        • super_sampling_ratio: Integer representing the ratio of the difference in size between the two image resolutions. Used for setting ratio of image subplots.
        • model_path: File path pointing to the folder where you want to save to model as well as generated samples.
        • dataset_path: File path pointing to the folder containing dataset you want to perform DLSS on.
        • save_path: File path pointing to the folder where you want to save generated predictions of the trained model.
        • png: Boolean flag, set to True if the data has PNGs to remove alpha layer from images.

  • Results

flip anime comparison

  • Generated Training Sample

Training

  • Generator Model Architecture

    • Using (5,5) Convolutional Kernels with 64 filters.
    • input_dimensions = (128,128,3) and output_dimensions = (256,256,3)
      Generator

  • Discriminator Model Architecture

    • Using (5,5) Convolutional Kernels with 64 filters.
    • input_dimensions = (128,128,3) and output_dimensions = (256,256,3)
      Discriminator
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].