TensorFlow Community Models

This repository provides a curated list of the GitHub repositories with machine learning models and implementations powered by TensorFlow 2.

Note: Contributing companies or individuals are responsible for maintaining their repositories.

Computer Vision

Image Recognition

Model	Paper	Features	Maintainer
DenseNet 169	Densely Connected Convolutional Networks	• FP32 Inference	Intel
Inception V3	[Rethinking the Inception Architecture
for Computer Vision](https://arxiv.org/pdf/1512.00567.pdf)	• Int8 Inference
• FP32 Inference	Intel
Inception V4	[Inception-v4, Inception-ResNet and the Impact
of Residual Connections on Learning](https://arxiv.org/pdf/1602.07261)	• Int8 Inference
• FP32 Inference	Intel
MobileNet V1	[MobileNets: Efficient Convolutional Neural Networks
for Mobile Vision Applications](https://arxiv.org/pdf/1704.04861)	• Int8 Inference
• FP32 Inference	Intel
ResNet 101	Deep Residual Learning for Image Recognition	• Int8 Inference
• FP32 Inference	Intel
ResNet 50	Deep Residual Learning for Image Recognition	• Int8 Inference
• FP32 Inference	Intel
ResNet 50v1.5	Deep Residual Learning for Image Recognition	• Int8 Inference
• FP32 Inference
• FP32 Training	Intel
EfficientNet v1 v2	EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks	• Automatic mixed precision
• Horovod Multi-GPU training (NCCL)
• Multi-node training on a Pyxis/Enroot Slurm cluster
• XLA	NVIDIA

Object Detection

Model	Paper	Features	Maintainer
R-FCN	[R-FCN: Object Detection
via Region-based Fully Convolutional Networks](https://arxiv.org/pdf/1605.06409)	• Int8 Inference
• FP32 Inference	Intel
SSD-MobileNet	[MobileNets: Efficient Convolutional Neural Networks
for Mobile Vision Applications](https://arxiv.org/pdf/1704.04861)	• Int8 Inference
• FP32 Inference	Intel
SSD-ResNet34	SSD: Single Shot MultiBox Detector	• Int8 Inference
• FP32 Inference
• FP32 Training	Intel

Segmentation

Model	Paper	Features	Maintainer
Mask R-CNN	Mask R-CNN	• Automatic Mixed Precision
• Multi-GPU training support with Horovod
• TensorRT	NVIDIA
U-Net Medical Image Segmentation	U-Net: Convolutional Networks for Biomedical Image Segmentation	• Automatic Mixed Precision
• Multi-GPU training support with Horovod
• TensorRT	NVIDIA

Natural Language Processing

Model	Paper	Features	Maintainer
BERT	[BERT: Pre-training of Deep Bidirectional Transformers
for Language Understanding](https://arxiv.org/pdf/1810.04805)	• FP32 Inference
• FP32 Training	Intel
BERT	BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding	• Horovod Multi-GPU
• Multi-node with Horovod and Pyxis/Enroot Slurm cluster
• XLA
• Automatic mixed precision
• LAMB	NVIDIA
ELECTRA	ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators	• Automatic Mixed Precision
• Multi-GPU training support with Horovod
• Multi-node training on a Pyxis/Enroot Slurm cluster	NVIDIA
GNMT	[Google’s Neural Machine Translation System:
Bridging the Gap between Human and Machine Translation](https://arxiv.org/pdf/1609.08144)	• FP32 Inference	Intel
Transformer-LT (Official)	Attention Is All You Need	• FP32 Inference	Intel
Transformer-LT (MLPerf)	Attention Is All You Need	• FP32 Training	Intel

Recommendation Systems

Model	Paper	Features	Maintainer
Wide & Deep	Wide & Deep Learning for Recommender Systems	• FP32 Inference
• FP32 Training	Intel
Wide & Deep	Wide & Deep Learning for Recommender Systems	• Automatic mixed precision
• Multi-GPU training support with Horovod
• XLA	NVIDIA
DLRM	Deep Learning Recommendation Model for Personalization and Recommendation Systems	• Automatic Mixed Precision
• Hybrid-parallel multiGPU training using Horovod all2all
• Multinode training for Pyxis/Enroot Slurm clusters
• XLA
• Criteo dataset preprocessing with Spark on GPU	NVIDIA

Contributions

If you want to contribute, please review the contribution guidelines.