Skip to content

Latest commit

 

History

History
 
 

3d-unet

MLPerf Inference Benchmarks for Medical Image 3D Segmentation

The chosen model is 3D-Unet in nnUnet performing BraTS 2019 brain tumor segmentation task.

Prerequisites

If you would like to run on NVIDIA GPU, you will need:

  • nvidia-docker
  • Any NVIDIA GPU supported by TensorFlow or PyTorch

Supported Models

model framework accuracy dataset model link model source precision notes
3D-Unet PyTorch mean = 0.85300 (whole tumor = 0.9141, tumor core = 0.8679, enhancing tumor = 0.7770) Fold 1 of BraTS 2019 Training Dataset from zenodo Trained in PyTorch using codes from nnUnet on Fold 0, Fold 2, Fold 3, and Fold 4 of BraTS 2019 Training Dataset. fp32
3D-Unet ONNX mean = 0.85300 (whole tumor = 0.9141, tumor core = 0.8679, enhancing tumor = 0.7770) Fold 1 of BraTS 2019 Training Dataset from zenodo Converted from the PyTorch model using script. fp32
3D-Unet Tensorflow mean = 0.85300 (whole tumor = 0.9141, tumor core = 0.8679, enhancing tumor = 0.7770) Fold 1 of BraTS 2019 Training Dataset from zenodo Converted from the ONNX model using script. fp32
3D-Unet OpenVINO mean = 0.85300 (whole tumor = 0.9141, tumor core = 0.8679, enhancing tumor = 0.7770) Fold 1 of BraTS 2019 Training Dataset from zenodo Converted from the ONNX model. fp32

Disclaimer

This benchmark app is a reference implementation that is not meant to be the fastest implementation possible.

Commands

Please download BraTS 2019 separately and unzip the dataset.

Please run the following commands:

  • export DOWNLOAD_DATA_DIR=<path/to/MICCAI_BraTS_2019_Data_Training>: point to location of downloaded BraTS 2019 Training dataset.
  • make setup: initialize submodule and download models.
  • make build_docker: build docker image.
  • make launch_docker: launch docker container with an interaction session.
  • make preprocess_data: preprocess the BraTS 2019 dataset.
  • python3 run.py --backend=[tf|pytorch|onnxruntime|ov] --scenario=[Offline|SingleStream|MultiStream|Server] [--accuracy] --model=[path/to/model_file(tf/onnx/OpenVINO only)]: run the harness inside the docker container. Performance or Accuracy results will be printed in console.
  • python3 accuracy-brats.py --log_file=<LOADGEN_LOG> --output_dtype=<DTYPE>: compute accuracy from a LoadGen accuracy JSON log file.

Details

  • SUT implementations are in ov_SUT.py, pytorch_SUT.py, onnxruntime_SUT.py, and tf_SUT.py. QSL implementation is in brats_QSL.py.
  • The script accuracy-brats.py parses LoadGen accuracy log, post-processes it, and computes the accuracy.
  • Preprocessing and evaluation (including post-processing) are not included in the timed path.
  • The input to the SUT is a volume of size [4, 224, 224, 160]. The output from SUT is a volume of size [4, 224, 224, 160] with predicted label logits for each voxel.

Calibration Set

The calibration set is the forty images listed in brats_cal_images_list.txt. They are randomly selected from Fold 0, Fold 2, Fold 3, and Fold 4 of BraTS 2019 Training Dataset.

License

Apache License 2.0