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 TensorFlow2
Resource
NVIDIA Deep Learning Examples
NVIDIA Deep Learning Examples
Mask R-CNN for TensorFlow2

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

Changelog

February 2021

  • Updated implementation to Tensorflow 2, using Keras API and Distributed strategy
  • ResNet50 checkpoint now is being downloaded from NVIDIA NGC
  • Training using batch size of 8 and 16 can result in unexpected hangs in DGX A100 80GB.

August 2020

  • Separated implementation for TensorFlow 1.1x and 2.x. New implementation is TF 1.1x version.
  • Recreated runtime part of the implementation.

June 2020

  • Updated accuracy tables with A100 results
  • Updated training and inference performance tables with A100 results

November 2019

  • Initial release

Known issues

  • Out of memory errors can occur when running training, V100, 8GPUs, BS4, FP32.
  • Errors can occur when running training with BS1.
  • The behavior of the model can be unstable when running with TensorFlow XLA enabled.