This article has compiled 57 of the most popular deep learning projects on GitHub (ranked by stars). Last updated: 2016.08.09
1.TensorFlow
Computes scalable machine learning problems using data flow diagrams
TensorFlow, Google’s second-generation machine learning system, performed twice as fast on some benchmarks as its first-generation DistBelief, according to the company.
TensorFlow’s extended support for built-in deep learning allows you to use TensorFlow for any computation that can be represented as a computational flow graph. Any gradient-based machine learning algorithm can benefit from the auto-differentiation of TensorFlow. It is also easy to express ideas in TensorFlow through the flexible Python interface.
2.Caffe
Stars: 11799
Caffe is a highly effective open source deep learning framework. It’s made up of expressions, speed and modularity.
3.Neural style
Stars: 10148
Torch implemented neural network algorithm.
Neural style is an algorithm that allows a machine to recreate an image by mimicking the style of an existing painting.
4.deepdream
Stars: 9042
Deep Dream, an image recognition tool
5.Keras
Stars: 7502
A deep learning library implemented by Python, including convolutional neural networks, recursive neural networks, etc. Runs on Theano and TensorFlow.
Keras is a minimalist, highly modular neural network library developed in Python (Python 2.7-3.5.) and can run on either TensorFlow or Theano platform. The good project aims to achieve rapid development of deep learning.
6.RocAlphaGo
Stars: 7170
An independent student-led project reconstructs DeepMind’s Nature 2016 paper, “Learning Go with Deep Neural Networks and Tree Search” (Nature 529, 484-489, 28 Jan 2016).
7.TensorFlow Models
Stars: 6671
A model developed based on TensorFlow
8.Neural Doodle
Stars: 6275
Turn doodles into elegant works of art with deep neural networks, generate seamless textures from photos, transform image styles, do instagram-based improvements, and more… There’s more! (Implementation of semantic style transfer)
9.CNTK
Stars: 5957
Deep learning toolkit. The CNTK toolkit from Microsoft is “crazier than anything we’ve ever seen”. This is partly thanks to CNTK’s ability to leverage graphics processing units (Gpus), which Microsoft claims to be the only company to unveil. When paired with the company’s networked GPU system, called Azure GPU Lab, it will be able to train deep neural networks to recognize speech, making the Cortana virtual assistant ten times faster than before.
10.TensorFlow Examples
Stars: 5872
TensorFlow tutorial and code examples for beginners, with notes and code explanations.
11.ConvNet JS
Stars: 5231
ConvNetJS is a neural network implemented in JavaScript, along with a browser-based demo.
12.Torch
Stars: 5133
Torch7, deep learning library.
Torch7 is a scientific computing framework that supports machine learning algorithms. It is easy to use and provides efficient algorithm implementation thanks to LuaJIT and an underlying C implementation.
13.OpenFace
Stars: 4855
Facial recognition based on deep learning network.
14.MXNet
Stars: 4685
A lightweight, portable, flexible distributed/mobile deep learning framework that supports Python, R, Julia, Scala, Go, Javascript and more.
MXNet is a deep learning framework designed for efficiency and flexibility. It allows you to mix symbolic and imperative programming to maximize efficiency and productivity. At its core is a dynamic dependency scheduler that can automatically parallel symbol and command operations. A graphical optimization layer that makes symbol execution fast and memory efficient. The library is portable, lightweight, and scalable to multiple Gpus and multiple machines.
15.Theano
Stars: 4286
Theano is a Python library that defines, optimizes, and emulates mathematical expression computations to efficiently solve multidimensional array computations.
16.Leaf
Stars: 4281
Open source machine intelligence framework for hackers.
17.Char RNN
Stars: 3820
A character-level language model for multi-level recursive neural networks, developed based on Torch.
18.Neural Talk
Stars: 3694
NeuralTalk is a Python+ Numpy project that uses multimodal recursive neural networks to describe images.
19.deeplearning4j
Stars: 3673
Java, Scala & Clojure deep learning tools based on Hadoop and Spark.
Deeplearning4j (DL4J for short) is the first commercial-grade open source distributed deep learning library written for Java and Scala. DL4J integrates with Hadoop and Spark and is designed for business environments rather than research tools. Skymind is DL4J’s business support organization.
Deeplearning4j’s advanced technology aims to plug and play, enabling rapid prototyping by non-researchers by using more presets without too much configuration. DL4J can also be customized at scale. DL4J is licensed under the Apache 2.0 license, and all derivative works based on DL4J belong to the authors of the derivative works.
20.TFLearn
Stars: 3368
Deep learning libraries, including the high-level TensorFlow interface.
21.TensorFlow Playground
Stars: 3352
Example of a neural network model.
22.OpenAI Gym
Stars: 3020
A toolkit for developing and comparing reinforcement learning algorithms.
23.Magenta
Stars: 2914
Magenta: Music and Art Generation and Machine Intelligence
A group of researchers from the Google Brain team has launched Project Magenta, whose main goal is to use machine learning to create art and compose music. Project Magenta uses the TensorFlow system, and the researchers have opened source their model and tools on GitHub.
Machine-generated music has been around for years, but it has always lacked a long narrative art, researchers say. Project Magenta tries to use stories as an important part of machine-generated music. Google released a DEMO (MP3) of the Magenta project’s results.
24.Colornet
Stars: 2798
Gray scale image is colored by neural network model.
25.Synaptic
Stars: 2666
Framework-free neural network library based on Node.js and browser.
26.Neural Talk 2
Stars: 2550
The image profile generation code developed by Torch runs on the GPU.
27.Image Analogies
Stars: 2540
Neural matching and fusion are used to generate similar patterns.
28.TensorFlow Tutorials
Stars: 2413
Tensorflow, from fundamentals to applications.
29.Lasagne
Stars: 2355
Lightweight function library based on Theano training and construction of neural network.
30.PyLearn2
Stars: 2153
Machine learning library based on Theano.
31.LISA-lab Deep Learning Tutorials
Stars: 2134
Deep learning tutorial notes and code. See the Wiki page for details.
32.Neon
Stars: 2121
Nervana™ Nervana™ is a fast, extensible and easy-to-use Python deep learning framework.
Neon is Nervana Systems’ deep learning software. Nervana’s software outperformed well-known deep learning tools, including Facebook’s own Torch7 and Nvidia’s cuDNN, according to benchmarks by a Facebook researcher.
33.Matlab Deep Learning Toolbox
Stars: 2032
Matlab/Octave deep learning toolbox. These include deep belief networks, automatic encoders, convolutional neural networks, convolutional automatic encoders, and Vanilla neural networks. There are getting started examples for each method.
34.Deep Learning Flappy Bird
Stars: 1721
Crack Flappy Bird with Deep reinforcement learning (Deep Q-learning).
35.dl-setup
Stars: 1607
Set up software instructions on the deep learning machine.
36.Chainer
Stars: 1573
A flexible neural network framework for deep learning.
Chainer is a framework for deep learning. Chainer Bridges the theoretical algorithms and practical applications of deep learning. It is characterized by being powerful, flexible and intuitive, and is considered a flexible framework for deep learning.
37.Neural Story Teller
Stars: 1514
A recursive neural network model for picture-telling.
38.DIGITS
Stars: 1353
Deep learning GPU training system.
39.Deep Jazz
Stars: 1229
Deep learning model for Generating Jazz based on Keras and Theano!
40.Tiny DNN
Stars: 1183
A deep learning framework that references only header files, has no dependencies and uses C ++ 11
41.Brainstorm
Stars: 1143
Fast, flexible and interesting neural networks.
42.dl-docker
Stars: 1044
An all-in-one Docker image for deep learning. Includes all popular DL frameworks (TensorFlow, Theano, Torch, Caffe, etc.).
43.Darknet
Stars: 937
C language version of the open source neural network.
44.Theano Tutorials
Stars: 904
Introduction to Machine learning based on Theano, from linear regression to convolutional Neural networks.
45.RNN Music Composition
Stars: 904
A recursive neural network tool for generating classical music.
46.Blocks
Stars: 866
Theano framework for constructing and training neural network models
47.TDB
Stars: 860
TensorFlow’s interactive, node debugging and visualization tool.
TensorDebugger (TDB) is a deep learning debugger that extends TensorFlow (Google’s deep learning framework) using breakpoints and computer graphics to visualize real-time data streams. In particular, TDB is a combination of a Python library and a Jupyter Notebook extension that builds on Google’s TensorFlow framework.
48.Scikit Neural Net
Stars: 849
Introduction to deep neural networks tools, sciKit-learn-like classifiers and regression models.
49.Veles
Stars: 760
Distributed Machine learning Platform (Python, CUDA, OpenCL)
VELES is a distributed deep learning application. Users only need to provide parameters and VELES can do the rest. VELES is written in Python, using OpenCL or CUDA, with flow-based programming. It is another TensorFlow developed by Samsung.
50.Deep Detect
Stars: 759
Deep learning interface and server based on C++11, bundled with Python and supported Caffe.
51.TensorFlow DeepQ
Stars: 759
Deep Q learning demo based on Google Tensorflow.
52.Caffe on Spark
Stars: 724
Caffe based on Spark.
Yahoo believes that deep learning should be in the same cluster as the existing data processing pipeline that supports feature engineering and traditional (non-deep) machine learning. We created CaffeOnSpark to enable deep learning training and testing to be embedded into the Spark application. CaffeOnSpark is designed as a Spark deep learning package.
53.Nolearn
Stars: 702
Abstract neural network library, famous Lasagne.
54.DCGAN TensorFlow
Stars: 568
Deep convolutional generative adversarial network based on Tensorflow.
55.MatConvNet
Stars: 479
MATLAB CNN computer vision application toolbox.
56.DeepCL
Stars: 413
OpenCL library for training deep convolutional neural network models.
57.Visual Search Server
Stars: 304
Visual search server. A simple visual search server implemented using TensorFlow, InceptionV3 model and AWS GPU instances.
The code implements two methods, a server that handles image searches and a simple indexer that extracts pool3 functionality. Nearest neighbor searches can be performed in an approximate manner using approximations (faster) or precise methods (slower).
Source: Top Deep Learning Projects