Resources

Open source repositories

• Model-to-Data approach for training deep learning without moving data: https://github.com/uw-biomedical-ml/oct-irf-train
• Online tool for collecting biomedical image segmentations: https://github.com/uw-biomedical-ml/segmentations
• HVF Progression Deep Learning Model Architecture: https://github.com/uw-biomedical-ml/hvfProgression
• Transforming between MRI modalities using deep learning:
  • https://github.com/uw-biomedical-ml/mri2mri
  • https://github.com/uw-biomedical-ml/mri-transformation
  • https://github.com/uw-biomedical-ml/mri2mri_notebooks
  • https://github.com/uw-biomedical-ml/mri-transformation
• Deep Learning based Intraretinal Fluid Segmentation on OCT: https://github.com/uw-biomedical-ml/irf-segmenter
• Deep Learning based OIR Quantification & Segmentation: https://github.com/uw-biomedical-ml/oir
• Tool for reading Heidelberg Spectralis imaging format: https://github.com/ayl/heyexReader

GPU capacities

• 46 x NVIDIA GPU cards for training and inference.
• Total computational capacity: 0.5 PetaFlops

Storage Pool

• 85 TB of stripped pool storage
• 12 of dual mirrored pools of 8 TB drives with 2 hot-spares (26 drives total)
• ZFS on Linux with online encryption and compression

Interconnect

• Mellanox Infiniband fiber optics with FDR (14 Gbs)

Logo

We would like to thank David Hoyt ( redanvil.design ) for the logo on our site.