Added
Highlights
- Framework for developing and deploying inference apps for streaming video data
- Supports MONAI Transforms for image transformations in the streaming pipeline
- Supports native inference (e.g using PyTorch) as well as Triton Inference Server model deployments
Sample Apps
- Ultrasound image segmentation using MONAI transforms on x86
- Ultrasound image segmentation using CUPy transformations for x86 and Clara Holoscan devices
- AJA video capture
data
This release contains example models and sample data for two streaming use cases -- ultrasound and endoscopy. Each dataset contains the models in the forms of pytorch jit, onyx and cagx-trt. Also included are licenses and sample videos, both the input video and ground truth segmentations.
- Ultrasound: for detecting spine scoliosis (full data is available [here](https://pocus.cs.queensu.ca/#collections))
- Endoscopy: CholecSeg8K data (full dataset available [here](https://www.kaggle.com/newslab/cholecseg8k)).