Skip to content

roboflow/jetson-utils

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

677 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

jetson-utils

C++/CUDA/Python multimedia utilities for NVIDIA Jetson:

cpp/ Various system & media utilities with C++ interfaces
    camera/ GStreamer-based camera capture (V4L2, MIPI CSI)
    codec/ GStreamer-based hardware video encoder/decoder
    display/ OpenGL window & rendering
    image/ Image loading & saving
    input/ Human Interface Devices (HID) from /dev/input
    network/ Sockets, IPv4/IPv6, WebRTC/RTSP server
    parsers/ Filesystem, CSV/JSON/XML parsing, command-line
    threads/ Multithreading, locks, and events
    video/ Video streaming interfaces
cuda/ CUDA image processing functions
docs/ Collection of Linux commands and links
python/ Python utilities, examples, and C++ bindings
scripts/ Standalone shell scripts in Bash or Python

JetPack 7 / Thor Support

This fork builds on JetPack 7.x (L4T r38/r39, CUDA 13) for Jetson Thor T5000/T4000 and Jetson Orin, verified against JetPack 7.2 (CUDA 13.2.1 / TensorRT 10 / Ubuntu 24.04 / Python 3.12):

  • On CUDA 13 the CUDA kernels are compiled for sm_87 (Orin) and sm_110 (Thor), and the library is built as C++17 (CUDA 13 dropped compilation for architectures below sm_75).
  • NVMM zero-copy memory remains disabled on JetPack 6 and newer (nvbuf_utils was removed from L4T) - video streams flow through system memory instead.
  • JetPack 4/5/6 builds are unchanged.

Documentation

Documentation for jetson-utils can be found here:

Assorted links and tips-and-tricks for Linux are kept under docs/

Building from Source (C++/CUDA)

This will build and install the C++/CUDA library (libjetson-utils.so) along with the Python extension module:

git clone https://github.com/dusty-nv/jetson-utils
cd jetson-utils
mkdir build
cd build
cmake ../
make -j$(nproc)
sudo make install
sudo ldconfig

If you're missing dependencies, run the jetson-inference/CMakePreBuild.sh script.

Install with Pip (Python only)

This will install the Python-native modules from python/jetson_utils which do not depend on the C++ extension bindings and get installed on top:

pip3 install -e /path/to/your/jetson-utils

Or directly from GitHub:

pip3 install git+https://github.com/dusty-nv/jetson-utils

About

C++/CUDA/Python multimedia utilities for NVIDIA Jetson

Resources

License

Stars

1 star

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • C++ 60.3%
  • C 24.2%
  • Cuda 6.8%
  • CMake 3.8%
  • Python 3.2%
  • Shell 1.7%