ContainersNVIDIADeepSeek-R1NVIDIALlama-3.1-Nemotron-70B-InstructNVIDIAPyTorch
CollectionsNVIDIAOmniverse Kit (FB)NVIDIADeepStream SDKNVIDIAOmniverse Kit App Streaming
ModelsNVIDIAStyleGAN3 pretrained modelsNVIDIAPeopleNetNVIDIATrafficCamNet
ResourcesNVIDIARiva Skills Quick Start
Helm ChartsNVIDIAGPU OperatorNVIDIANVIDIA NIM Operator
NVIDIA Deep Learning Examples
NVIDIA Deep Learning Examples
Mask R-CNN for PyTorch
Resource
NVIDIA Deep Learning Examples
NVIDIA Deep Learning Examples
Mask R-CNN for PyTorch

Mask R-CNN is a convolution based network for object instance segmentation. This implementation provides 1.3x faster training while maintaining target accuracy.

Changelog

May 2022

  • Update container to 21.12
  • Add native NHWC, native DDP
  • Replace SGD with FusedSGD
  • Use cuda streams to prefetch dataloader
  • Hybrid dataloader
  • Use new training recipe

October 2021

  • Replace APEX AMP with PyTorch native AMP
  • Use opencv-python version 4.4.0.42

July 2021

  • Update container
  • Use native AMP
  • Update dataset to coco 2017

June 2020

  • Updated accuracy and performance tables to include A100 results

September 2019

  • Updates for PyTorch 1.2
  • Jupyter notebooks added

July 2019

  • Update AMP to new API
  • Update README
  • Download support from torch hub
  • Update default test batch size to 1/gpu

March 2019

  • Initial release

Known Issues

Currently V100 32GB + FP32 + NHWC + Hybrid dataloader causes a slowdown. So for all V100 32GB FP32 runs hybrid dataloader and NHWC should be disabled.