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

This repo contains the source code in my personal column (https://zhuanlan.zhihu.com/zhaoyeyu), implemented using Python 3.6. Including Natural Language Processing and Computer Vision projects, such as text generation, machine translation, deep convolution GAN and other actual combat code.

Stars: ✭ 3,307 (+1868.45%)

Mutual labels: convolutional-neural-networks, autoencoder

Motion Sense

MotionSense Dataset for Human Activity and Attribute Recognition ( time-series data generated by smartphone's sensors: accelerometer and gyroscope)

Stars: ✭ 159 (-5.36%)

Mutual labels: convolutional-neural-networks, autoencoder

Neurec

Next RecSys Library

Stars: ✭ 731 (+335.12%)

Mutual labels: convolutional-neural-networks, autoencoder

Niftynet

[unmaintained] An open-source convolutional neural networks platform for research in medical image analysis and image-guided therapy

Stars: ✭ 1,276 (+659.52%)

Mutual labels: convolutional-neural-networks, autoencoder

Sign Language Interpreter Using Deep Learning

A sign language interpreter using live video feed from the camera.

Stars: ✭ 157 (-6.55%)

Mutual labels: convolutional-neural-networks

Tf Adnet Tracking

Deep Object Tracking Implementation in Tensorflow for 'Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning(CVPR 2017)'

Stars: ✭ 162 (-3.57%)

Mutual labels: convolutional-neural-networks

Human Action Recognition With Keras

Keras implementation of Human Action Recognition for the data set State Farm Distracted Driver Detection (Kaggle)

Stars: ✭ 156 (-7.14%)

Mutual labels: convolutional-neural-networks

Sparse Winograd Cnn

Efficient Sparse-Winograd Convolutional Neural Networks (ICLR 2018)

Stars: ✭ 156 (-7.14%)

Mutual labels: convolutional-neural-networks

Vdcnn

Implementation of Very Deep Convolutional Neural Network for Text Classification

Stars: ✭ 158 (-5.95%)

Mutual labels: convolutional-neural-networks

Sequitur

Library of autoencoders for sequential data

Stars: ✭ 162 (-3.57%)

Mutual labels: autoencoder

Brain.js

brain.js is a GPU accelerated library for Neural Networks written in JavaScript.

Stars: ✭ 12,358 (+7255.95%)

Mutual labels: convolutional-neural-networks

A Journey Into Convolutional Neural Network Visualization

A journey into Convolutional Neural Network visualization

Stars: ✭ 165 (-1.79%)

Mutual labels: convolutional-neural-networks

Imagenet

TensorFlow implementation of AlexNet and its training and testing on ImageNet ILSVRC 2012 dataset

Stars: ✭ 155 (-7.74%)

Mutual labels: convolutional-neural-networks

Tensorflow Cnn Finetune

Finetuning AlexNet, VGGNet and ResNet with TensorFlow

Stars: ✭ 166 (-1.19%)

Mutual labels: convolutional-neural-networks

Visual Interaction Networks Pytorch

This's an implementation of deepmind Visual Interaction Networks paper using pytorch

Stars: ✭ 164 (-2.38%)

Mutual labels: convolutional-neural-networks

Danmf

A sparsity aware implementation of "Deep Autoencoder-like Nonnegative Matrix Factorization for Community Detection" (CIKM 2018).

Stars: ✭ 161 (-4.17%)

Mutual labels: autoencoder

Iresnet

Improved Residual Networks (https://arxiv.org/pdf/2004.04989.pdf)

Stars: ✭ 163 (-2.98%)

Mutual labels: convolutional-neural-networks

View All Similar Projects ➔

PyTorch Deep Image Prior

An implementation of image reconstruction methods from Deep Image Prior (Ulyanov et al., 2017) in PyTorch.

The point of the paper is to execute some common image manipulation tasks using neural networks untrained on data prior to use.

Architectures differ from those used in the actual paper. The authors use some specific networks for specific tasks. This repo uses a couple of alternative architectures to produce similar results. One where upsampling is done on the basis of pixel shuffling, and the other using transposed convolutions. Pixel shuffling results in some hotspots that do not disappear with further training.

Requirements

Python3 with PyTorch, torchvision and NumPy. CUDA and cuDNN are optional (settable within the script in a self-explanatory way) but strongly recommended.

To use

It's relatively easy to play around with the settings from within the scripts. To reproduce the results in the repo, do the following.

Make a directory to hold the network output:

mkdir output

Generate output images with:

python3 deep_image_prior.py

Consolidate output images into a training gif and sample some actual data with:

python3 parse_ec2_results.py

Results

Note that the images here have been downsampled for formatting sensibly in the README. Full sized samples are in the repo if you would like to have a closer look.

Training was done over 25k iterations on an Amazon GPU instance. Takes roughly an hour on 512x512 images.

Upsampling with transposed convolutions:

Note the grid like speckles during training. These are caused by convolutional kernels overlapping with one another during upsampling.

Ground truth	Input	Output	Training

Upsampling with pixel shuffling:

No speckles, however there is a hotspot (in the out of focus region towards the bunny's behind) that becomes a black spot. The appearance of these hotspots seems commonplace through both architectures, but the extra smoothness given by the convolution transpose layers repairs these more effectively.

Ground truth	Input	Output	Training

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].

Stars: ✭ 168

Visit Git Page 🔗Visit User Page 🔗Visit Issues Page (0) 🔗