Open-source deep-learning software
2663 open-source libraries, tools and applications tagged deep-learning, ranked by how closely deep-learning is their primary purpose, then by adoption and maintenance health. Each entry links to a page with sources, use cases and when to choose or avoid it.
2663 products, primary matches first, then adoption-weighted; health v2 shown.
| Product | Health v2 | Stars | Maturity |
|---|---|---|---|
| lizhihao6/Sparc3D Sparc3D is the official implementation of a research framework for high-resolution 3D shape modeling, combining a sparse deformable marchin… | 28 | 1359 | experimental |
| Tencent-Hunyuan/Hunyuan3D-WorldClaw WorldClaw is a research framework from Tencent Hunyuan for agentic, large-scale 3D open-world generation, released alongside an arXiv paper… | 53 | 1334 | experimental |
| lucidrains/naturalspeech2-pytorch A PyTorch implementation of NaturalSpeech 2, a zero-shot text-to-speech and singing synthesizer that combines a neural audio codec with a l… | 21 | 1334 | experimental |
| google/style-aligned Official research code for 'Style Aligned Image Generation via Shared Attention', implementing style-consistent image generation with diffu… | 10 | 1315 | experimental |
| DLYuanGod/TinyGPT-V TinyGPT-V is an efficient multimodal large language model built on small backbones (Phi-2 2.7B), combining vision and language capabilities… | 53 | 1314 | experimental |
| Jamie-Stirling/RetNet A minimal, pure PyTorch implementation of the Retentive Network (RetNet) architecture proposed as a successor to Transformers for large lan… | 29 | 1212 | experimental |
| tspeterkim/flash-attention-minimal A minimal ~100-line CUDA re-implementation of the Flash Attention forward pass, loadable as a PyTorch extension. It is designed as an educa… | 26 | 1192 | experimental |
| facebookresearch/shumai Shumai is a fast, differentiable tensor library for TypeScript and JavaScript built on Bun and Flashlight (ArrayFire backend). It provides … | 32 | 1171 | experimental |
| Ligo-Biosciences/AlphaFold3 An open-source Python implementation of AlphaFold3, DeepMind's biomolecular structure prediction model, including the full model architectu… | 27 | 1093 | experimental |
| volcengine/veScale veScale is a PyTorch distributed training library from ByteDance for hyperscale training of large language models and reinforcement learnin… | 54 | 1041 | experimental |
| HumanMLLM/R1-Omni R1-Omni is a research project applying Reinforcement Learning with Verifiable Reward (RLVR) to an omni-multimodal large language model for … | 22 | 1025 | experimental |
| orca-wm/Orca Orca is a general world foundation model from BAAI centered on Next-State-Prediction, learning a unified world latent space from visual and… | 54 | 1021 | experimental |
| meta-llama/llama Meta's official repository containing inference code and download scripts for Llama 2 large language models (7B to 70B parameters). The rep… | 31 | 59632 | abandoned |
| mozilla/DeepSpeech DeepSpeech is an open-source, offline speech-to-text engine based on Baidu's Deep Speech research paper and implemented with TensorFlow. It… | 10 | 26774 | abandoned |
| facebookresearch/Detectron Facebook AI Research's Python software system implementing state-of-the-art object detection algorithms such as Mask R-CNN, RetinaNet, and … | 10 | 26352 | abandoned |
| openai/gpt-2 Official OpenAI repository containing code and pretrained model weights for GPT-2, from the paper 'Language Models are Unsupervised Multita… | 10 | 25009 | abandoned |
| QwenLM/Qwen The official repository for Qwen (Tongyi Qianwen), Alibaba Cloud's family of pretrained and chat large language models, with weights on Hug… | 57 | 21858 | abandoned |
| apache/mxnet Apache MXNet is a deep learning framework offering a hybrid front-end that mixes imperative (Gluon) and symbolic programming, with a dynami… | 10 | 20804 | abandoned |
| Magenta Magenta is an open‑source Python library that provides TensorFlow models and tools for AI‑driven music and art creation. It includes pre‑tr… | 10 | 19787 | abandoned |
| Magenta Magenta is a Google Brain research project and Python TensorFlow library for generating music, images, and drawings with machine learning. … | 10 | 19787 | abandoned |
| jantic/DeOldify DeOldify is a deep learning project for colorizing and restoring old black-and-white images and videos, with artistic and stable model vari… | 10 | 18463 | abandoned |
| jcjohnson/neural-style A Torch (Lua) implementation of the Gatys et al. neural style transfer algorithm, which combines the content of one image with the artistic… | 32 | 18269 | abandoned |
| microsoft/CNTK The Microsoft Cognitive Toolkit (CNTK) is an open-source deep-learning toolkit that describes neural networks as directed computational gra… | 10 | 17584 | abandoned |
| horovod/horovod Horovod is a distributed deep learning training framework for TensorFlow, Keras, PyTorch, and Apache MXNet, originally developed at Uber. I… | 10 | 14680 | abandoned |
| alexjc/neural-enhance Neural Enhance is a Python command-line tool that upscales images 2x or 4x using deep learning super-resolution models. It ships with pre-t… | 10 | 11865 | abandoned |
| karpathy/convnetjs ConvNetJS is a JavaScript library for defining and training neural networks, including convolutional networks and an experimental Deep Q Le… | 23 | 11208 | abandoned |
| facebookresearch/maskrcnn-benchmark A fast, modular PyTorch reference implementation of instance segmentation and object detection algorithms including Mask R-CNN, Faster R-CN… | 10 | 9361 | abandoned |
| Torch7 Torch7 is a scientific computing framework with a fast multi-dimensional tensor library and math operations, originally built for Lua with … | 35 | 9143 | abandoned |
| TensorFlow.js tfjs-core was the core WebGL-accelerated tensor computation library behind TensorFlow.js, providing linear algebra operations and automatic… | 10 | 8425 | abandoned |
| facebookarchive/caffe2 Caffe2 is a lightweight, modular, and scalable deep learning framework built on the original Caffe, with both Python and C++ APIs. It has b… | 10 | 8370 | abandoned |
| rbgirshick/py-faster-rcnn A Python reimplementation of the Faster R-CNN object detection model built on a fork of Fast R-CNN and Caffe. The repository is officially … | 32 | 8289 | abandoned |
| openai/jukebox OpenAI's research code and pretrained model weights for Jukebox, a generative neural network that produces music with vocals as raw audio f… | 10 | 8028 | abandoned |
| harthur/brain A simple feed-forward neural network library written in JavaScript, usable in Node.js and the browser. It supports training on input/output… | 10 | 7984 | abandoned |
| jwyang/faster-rcnn.pytorch A pure PyTorch implementation of Faster R-CNN for object detection, supporting multi-image batch training and multi-GPU training with sever… | 32 | 7859 | abandoned |
| fchollet/deep-learning-models A deprecated collection of Keras code and pre-trained weights for popular deep learning image classification models such as VGG16, VGG19, R… | 23 | 7351 | abandoned |
| facebookresearch/DensePose DensePose is a research library from Facebook AI that maps all human pixels in 2D RGB images to a 3D surface-based model of the human body … | 10 | 7261 | abandoned |
| carpedm20/DCGAN-tensorflow A TensorFlow implementation of Deep Convolutional Generative Adversarial Networks (DCGAN) for generating images such as faces and digits. I… | 32 | 7178 | abandoned |
| facebookresearch/pytext PyText is a deep-learning-based NLP modeling framework built on PyTorch, providing extensible abstractions for text classifiers, sequence t… | 10 | 6291 | abandoned |
| Swift for TensorFlow Swift for TensorFlow was an experimental next-generation machine learning platform integrating differentiable programming directly into the… | 10 | 6121 | abandoned |
| Swift AI Swift AI is a high-performance deep learning library written entirely in Swift, offering a flexible fully-connected neural network optimize… | 23 | 6036 | abandoned |
| tiny-dnn/tiny-dnn tiny-dnn is a header-only, dependency-free deep learning framework written in C++14, designed for training and running neural networks on r… | 23 | 6028 | abandoned |
| yahoo/open_nsfw A Python library from Yahoo that runs a Caffe deep neural network to classify images as Not Suitable for Work (NSFW), outputting a probabil… | 10 | 6014 | abandoned |
| salesforce/BLIP PyTorch implementation of BLIP, a vision-language pre-training model for unified image understanding and text generation tasks. The reposit… | 10 | 5708 | abandoned |
| google/seq2seq A general-purpose encoder-decoder (seq2seq) framework built on TensorFlow, supporting tasks like machine translation, text summarization, c… | 10 | 5619 | abandoned |
| karpathy/neuraltalk2 NeuralTalk2 is a Torch (Lua) implementation of image captioning that uses a CNN (VGGNet) followed by an RNN language model to generate text… | 32 | 5600 | abandoned |
| Leaf Leaf is an open-source machine learning framework written in Rust for building classical, deep, or hybrid ML applications, inspired by Tens… | 32 | 5540 | abandoned |
| karpathy/neuraltalk NeuralTalk is a Python+numpy implementation of Multimodal Recurrent Neural Networks that generate natural-language descriptions of images. … | 32 | 5508 | abandoned |
| david-gpu/srez A deep learning project that performs 4x image super-resolution on 16x16 images using a DCGAN-based architecture with ResNet generator modu… | 10 | 5266 | abandoned |
| ashnkumar/sketch-code A Keras/TensorFlow deep learning model that converts hand-drawn website wireframes into HTML code using an image captioning architecture (C… | 72 | 5141 | abandoned |
| AutoGPTQ/AutoGPTQ AutoGPTQ is a Python library for quantizing large language models using the GPTQ weight-only quantization algorithm, with user-friendly API… | 10 | 5065 | abandoned |
| transcranial/keras-js A JavaScript library for running trained Keras deep learning models directly in the browser, with GPU acceleration via WebGL 2. It also sup… | 23 | 4964 | abandoned |
| DenseNet Reference implementation of DenseNet (Densely Connected Convolutional Networks), the CVPR 2017 Best Paper Award-winning CNN architecture, w… | 32 | 4872 | abandoned |
| idealo/image-super-resolution A Python library providing Keras implementations of Residual Dense and Adversarial Networks for single image super-resolution, including pr… | 10 | 4815 | abandoned |
| amazon-archives/amazon-dsstne DSSTNE (Deep Scalable Sparse Tensor Network Engine) is a C++ GPU library from Amazon for training and deploying deep learning recommendatio… | 10 | 4390 | abandoned |
| RetinaNet (Keras/TF implementation) A Keras/TensorFlow implementation of the RetinaNet object detection model with focal loss, supporting training and inference on custom data… | 23 | 4382 | abandoned |
| janhuenermann/neurojs neurojs is a JavaScript deep learning framework for the browser with a focus on reinforcement learning, supporting deep Q-networks and acto… | 32 | 4367 | abandoned |
| NVIDIA/DIGITS DIGITS is a web application for training deep learning models on GPUs, supporting frameworks like Caffe, Torch, and TensorFlow with a brows… | 10 | 4176 | abandoned |
| goodfeli/adversarial Reference implementation of the original 2014 'Generative Adversarial Networks' paper by Goodfellow et al., containing training code, YAML … | 32 | 4087 | abandoned |
| google-deepmind/learning-to-learn A TensorFlow implementation of 'Learning to learn by gradient descent by gradient descent' (DeepMind, arXiv:1606.04474), which trains LSTM-… | 32 | 4072 | abandoned |
| junyanz/iGAN iGAN is a research application implementing interactive image generation with generative adversarial networks, letting users draw strokes t… | 32 | 4012 | abandoned |
| buriburisuri/speech-to-text-wavenet A TensorFlow implementation of end-to-end English speech recognition based on DeepMind's WaveNet architecture, trained with CTC loss on sen… | 32 | 4003 | abandoned |
| udacity/self-driving-car-sim A Unity-based self-driving car simulator built for Udacity's Self-Driving Car Nanodegree, used to train cars to navigate road courses with … | 10 | 3984 | abandoned |
| rasmusbergpalm/DeepLearnToolbox A Matlab/Octave toolbox implementing classic deep learning models including Deep Belief Nets, Stacked Autoencoders, Convolutional Neural Ne… | 32 | 3869 | abandoned |
| NervanaSystems/neon neon was Intel Nervana's reference deep learning framework in Python, focused on best-in-class training performance on CPUs, GPUs, and Nerv… | 10 | 3860 | abandoned |
| facebookresearch/fairseq-lua Fairseq-lua is Facebook AI Research's sequence-to-sequence learning toolkit for the Torch framework, focused on neural machine translation … | 10 | 3725 | abandoned |
| fo40225/tensorflow-windows-wheel A repository of prebuilt TensorFlow wheel binaries for 64-bit Windows, including CPU-only and CUDA/cuDNN GPU builds with AVX2 and SSE2 (non… | 32 | 3665 | abandoned |
| endernewton/tf-faster-rcnn A TensorFlow implementation of the Faster R-CNN object detection framework, supporting VGG16, ResNet, and MobileNet backbones trained and e… | 23 | 3649 | abandoned |
| apple/tensorflow_macos A pre-release fork of TensorFlow and TensorFlow Addons for macOS 11.0+ that provides hardware acceleration on Apple Silicon and Intel Macs … | 10 | 3643 | abandoned |
| pavelgonchar/colornet Colornet is a Python/TensorFlow neural network implementation that colorizes grayscale images, based on VGG16 features and YUV color-channe… | 32 | 3550 | abandoned |
| Newmu/dcgan_code Reference implementation of the DCGAN paper (Unsupervised Representation Learning with Deep Convolutional GANs) in Python. It trains convol… | 32 | 3529 | abandoned |
| rbgirshick/fast-rcnn Fast R-CNN is Ross Girshick's ICCV 2015 framework for fast object detection with deep convolutional networks, written in Python and C++/Caf… | 32 | 3462 | abandoned |
| tensorforce/tensorforce Tensorforce is an open-source deep reinforcement learning framework built on TensorFlow, emphasizing modular, configurable library design f… | 64 | 3305 | abandoned |
| openai/glow Official research code from OpenAI for reproducing the 'Glow: Generative Flow with Invertible 1x1 Convolutions' paper, a flow-based generat… | 10 | 3181 | abandoned |
| tensorflow/skflow A simplified Scikit Learn-style interface for TensorFlow deep learning, providing estimators for neural networks. It was merged into Tensor… | 10 | 3165 | abandoned |
| LeeJunHyun/Image_Segmentation A PyTorch implementation of four U-Net variants for image segmentation: U-Net, R2U-Net, Attention U-Net, and Attention R2U-Net. It includes… | 32 | 3102 | abandoned |
| mistralai/mistral-finetune A lightweight Python codebase from Mistral AI for memory-efficient LoRA fine-tuning of Mistral's language models, optimized for single-node… | 10 | 3094 | abandoned |
| CharlesShang/FastMaskRCNN A TensorFlow implementation of Mask R-CNN for instance segmentation, reproducing the paper by Kaiming He et al. It includes ROIAlign, a COC… | 32 | 3082 | abandoned |
| raghakot/keras-vis keras-vis is a high-level Python toolkit for visualizing and debugging trained Keras neural network models. It supports activation maximiza… | 23 | 2992 | abandoned |
| ryankiros/neural-storyteller A recurrent neural network research project that generates short stories about images by combining skip-thought vectors, visual-semantic em… | 32 | 2959 | abandoned |
| IntelLabs/nlp-architect NLP Architect is an open-source Python library from Intel AI Lab for exploring state-of-the-art deep learning topologies and optimization t… | 10 | 2931 | abandoned |
| jisungk/deepjazz deepjazz is a Python application that uses a two-layer LSTM built with Keras and Theano to generate jazz music from a given MIDI file. It w… | 61 | 2895 | abandoned |
| jetpacapp/DeepBeliefSDK Jetpac's Deep Belief SDK, a cross-platform image recognition framework implementing the AlexNet convolutional neural network architecture, … | 32 | 2851 | abandoned |
| isl-org/ZoeDepth ZoeDepth is a PyTorch library implementing metric depth estimation from a single image, combining relative and metric depth approaches with… | 10 | 2838 | abandoned |
| ShaoqingRen/faster_rcnn A MATLAB re-implementation of Faster R-CNN, a deep-learning object detection framework combining a Region Proposal Network with a detection… | 32 | 2835 | abandoned |
| zzw922cn/Automatic_Speech_Recognition An end-to-end automatic speech recognition system implemented in TensorFlow, supporting Mandarin and English with models like DeepSpeech2, … | 32 | 2833 | abandoned |
| Neural Network Libraries (NNabla) Sony's Neural Network Libraries (nnabla) is a deep learning framework with a Python API over a C++11 core, designed for research, developme… | 10 | 2773 | abandoned |
| lisa-lab/pylearn2 Pylearn2 is a Python machine learning research library built on top of Theano, developed by the LISA lab at Université de Montréal. It prov… | 32 | 2765 | abandoned |
| kaonashi-tyc/zi2zi zi2zi is a TensorFlow implementation of a conditional GAN (based on pix2pix) for learning and transferring Chinese and East Asian character… | 32 | 2752 | abandoned |
| allanzelener/YAD2K YAD2K is a 90% Keras / 10% TensorFlow implementation and converter for YOLO_v2 object detection models. It converts Darknet cfg and weight … | 32 | 2728 | abandoned |
| NVlabs/MUNIT Official PyTorch implementation of MUNIT (ECCV 2018), a GAN-based model for multimodal unsupervised image-to-image translation. The reposit… | 23 | 2706 | abandoned |
| intel/BigDL BigDL is Intel's distributed deep learning library that scales TensorFlow, Keras, and PyTorch workloads on Apache Spark, Flink, and Ray, wi… | 10 | 2699 | abandoned |
| propelml/propel Propel was a JavaScript library for differentiable programming, providing tensor operations and automatic differentiation for scientific co… | 10 | 2690 | abandoned |
| ryanjay0/miles-deep Miles Deep is a C++ deep learning application built on Caffe that classifies each second of a pornographic video into six sexual act catego… | 23 | 2662 | abandoned |
| sherjilozair/char-rnn-tensorflow A Python/TensorFlow implementation of multi-layer recurrent neural networks (LSTM) for character-level language modeling, inspired by Andre… | 32 | 2658 | abandoned |
| coqui-ai/STT Coqui STT is an open-source deep learning toolkit for training and deploying speech-to-text models, built on TensorFlow with bindings for m… | 23 | 2605 | abandoned |
| devsisters/DQN-tensorflow A TensorFlow implementation of the Deep Q-Network (DQN) algorithm from the DeepMind paper 'Human-Level Control through Deep Reinforcement L… | 32 | 2586 | abandoned |
| rbgirshick/rcnn The original R-CNN (Region-based Convolutional Neural Networks) object detection system from UC Berkeley, released as research code accompa… | 23 | 2416 | abandoned |
| karanchahal/DoodleMaster Doodle Master is a proof-of-concept AI tool that turns hand-drawn UI mockups into real HTML/CSS code using a CNN classifier for detecting U… | 32 | 2408 | abandoned |
| tryolabs/luminoth Luminoth is an open-source Python toolkit for computer vision built on TensorFlow and Sonnet, focused on object detection with Faster R-CNN… | 23 | 2406 | abandoned |
| kaishengtai/neuralart A Torch7 (Lua) implementation of the paper 'A Neural Algorithm of Artistic Style' for neural style transfer, applying the style of one imag… | 32 | 2394 | abandoned |