Open Science

Open Code & Datasets

We believe in reproducible AI research. Access our publicly released benchmark datasets, pre-trained model weights, and open-source code repositories.

Benchmark Dataset DICOM / NIfTI (120 GB)

MedVis3D-Bench: 5,000 Multi-Contrast Brain MRI Scans

A benchmark dataset of 5,000 anonymized multi-contrast (T1w, T2w, FLAIR) brain MRI scans curated for zero-shot 3D segmentation evaluation.

License: CC-BY 4.0 Access Data →

Model Weights PyTorch / ONNX (4.2 GB)

OmniVision-3D Pre-trained Checkpoints

PyTorch model weights pre-trained on 150,000 unannotated volumetric CT & MRI scans for zero-shot medical transfer learning.

License: MIT License Access Data →

Benchmark Dataset HDF5 / SVS (450 GB)

PathoBiomarker-50K Whole Slide Biopsies

Multi-center digital pathology corpus containing 50,000 gigapixel whole-slide tissue biopsies with clinical outcome labels.

License: CC-BY-NC 4.0 Access Data →

Model Weights TensorRT / CoreML (45 MB)

CardioEcho-Realtime Point-of-Care Model

Lightweight 60 FPS mobile neural network for real-time ejection fraction quantification on point-of-care ultrasound devices.

License: Apache 2.0 Access Data →