Legacy example · C++ · Python · Version 1.0
Distributed H.264 Endoscopy Tool Tracking
| Authors | Holoscan Team (NVIDIA) |
|---|---|
| Supported platforms | x86_64, aarch64 |
| Language | C++, Python |
| Version | 1.0 |
| Compatibility | HSDK 2.6 |
| Contribution level | Level 1 · Very high-quality code |
| Source | HoloHub |
Legacy HoloHub application
This application comes from HoloHub and was last tested with Holoscan SDK (HSDK) 2.6. It runs with the HoloHub CLI, ./holohub.
This application is similar to the H.264 Endoscopy Tool Tracking application, but this distributed version divides the application into three fragments:
- Video Input: get video input from a pre-recorded video file.
- Inference: run the inference using LSTM and run the post-processing script.
- Visualization: display input video and inference results.
Requirements
This application is configured to use H.264 elementary stream from endoscopy sample data as input.
Data
📦️ (NGC) Sample App Data for AI-based Endoscopy Tool Tracking
The data is automatically downloaded when building the application.
Building and Running H.264 Endoscopy Tool Tracking Application
- Building and running the application from the top level Holohub directory:
C++ (build and run)
# Start the application with all three fragments
./holohub run ucx_h264_endoscopy_tool_tracking --language=cpp
# Use the following commands to run the same application three processes:
# Start the application with the video_in fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=cpp --run-args="--driver --worker --fragments video_in --address :10000 --worker-address :10001"
# Start the application with the inference fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=cpp --run-args="--worker --fragments inference --address :10000 --worker-address :10002"
# Start the application with the visualization fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=cpp --run-args="--worker --fragments viz --address :10000 --worker-address :10003"
Python (build and run)
# Start the application with all three fragments
./holohub run ucx_h264_endoscopy_tool_tracking --language=python
# Use the following commands to run the same application three processes:
# Start the application with the video_in fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=python --run-args="--driver --worker --fragments video_in --address :10000 --worker-address :10001"
# Start the application with the inference fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=python --run-args="--worker --fragments inference --address :10000 --worker-address :10002"
# Start the application with the visualization fragment
./holohub run ucx_h264_endoscopy_tool_tracking --language=python --run-args="--worker --fragments viz --address :10000 --worker-address :10003"
Important: on aarch64, applications also need tegra folder mounted inside the container and
the LD_LIBRARY_PATH environment variable should be updated to include
tegra folder path.
Open and edit the Dockerfile and uncomment line 66:
# Uncomment the following line for aarch64 support
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/aarch64-linux-gnu/tegra/