Useful links
Nvidia Isaac ROS 2.0
Isaac ROS 2.0 is a software stack for ROS2 (Apache 2.0) set of libraries.
It is made up of open source (Apache 2.0) packages on GitHub
https://developer.nvidia.com/embedded/jetpack
https://nvidia-isaac-ros.github.io/performance/index.html
Object Detection
Darknet is a Neural Network framework with You Only Look Once (YOLO) object detection algorithms, available under various open source licenses.
- Performance measurement of on-device YOLOv4 deployed with Nvidia DeepStream SDK manages 13 FPS average
3D Flash LiDAR Object Detection and Tracking on Edge Hardware
- 68 FPS average on Orin Nano with cuNN
Drone stuff
Design and Flight Demonstration of a Quadrotor for Urban Mapping and Target Tracking Research
- Following a moving GPS target, scene reconstruction with 5 cameras
Analysis of Deep Feature Matching Algorithms in UAV Visual Localization
CNNs
Hierarchical Feature Pooling Transformer for Efficient UAV Object Tracking
Detecting Violent Behaviour on Edge Using Convolutional Neural Networks
SLAM
High-Speed Stereo Visual SLAM for Low-Powered Computing Devices
- Jetson-Slam (GPLv3), for original Jetson devices. 60 FPS on Jetson-NX.