Operators

These are the operators included as part of the Holoscan SDK:

Video acquisition / transmission (video_io)

The vendor-agnostic bases **VideoAcquisitionOperator** and **VideoTransmissionOperator**, capability types, and the **video_io** registry live in the shared library built as CMake target **op_video_io** (import target **holoscan::ops::video_io**). They are on by default when you build the Holoscan SDK from source; no extra CMake flag is required.

• Configure / build (from the public/ tree or via ./run build; see DEVELOP.md): a normal SDK build produces **libholoscan_op_video_io.so** (name may vary by platform) and installs headers under include/holoscan/operators/video_io/.

• Link your app / operator: target_link_libraries(<your_target> PRIVATE holoscan::core holoscan::ops::video_io) (or PUBLIC if you expose the bases in your API).

There is no separate flag for “transmission only”: acquisition and transmission share **op_video_io**.

• bayer_demosaic: perform color filter array (CFA) interpolation for 1-channel inputs of 8 or 16-bit unsigned integer and outputs an RGB or RGBA image

• format_converter: provides common video or tensor operations in inference pipelines to change datatypes, resize images, reorder channels, and normalize and scale values.

• gxf_codelet: Provides a generic import interface for a GXF Codelet.

• holoviz: handles compositing, blending, and visualization of RGB or RGBA images, masks, geometric primitives, text and depth maps

• inference: performs AI inference using APIs from HoloInfer module.

• inference_processor: performs processing of data using APIs from HoloInfer module. In the current release, a limited set of operations are supported on CPU.

• ping_rx: “receive” and log an int value

• ping_tx: “transmit” an int value

• segmentation_postprocessor: generic AI postprocessing operator

• tensor_rt *(deprecated)*: perform AI inference with TensorRT

• test_ops : operators intended primarily for use in test cases. Currently contains a DataTypeTxTestOp which is a source operator which can be configured to emit data of many different basic C++ types. It also contains a DataTypeRxTestOp which receives via std::any and prints the type name of the received type. These are being used in Python test applications to test automated conversion from C++ types to/from equivalent Python types on receive or emit of various data types from a native Python operator. It also includes PoseTreeManagerLookupOp for validating fragment service lookup from C++ when a service is registered in Python.

• v4l2_video_capture: V4L2 Video Capture

• video_io: vendor-agnostic base operators (VideoAcquisitionOperator, VideoTransmissionOperator) for multi-stream video capture and playout. Hardware vendors subclass these to provide SDK-native integration. See the Vendor Implementation Guide for details.

• video_stream_recorder: write a video stream output as .gxf_entities + .gxf_index files on disk

• video_stream_replayer: read .gxf_entities + .gxf_index files on disk as a video stream input