An inference operator for TensorFlow-based object detection models. Designed for modularity and ease of use.
Only supported on Ubuntu 18.04 at this time. You'll need the following dependencies:
- NVIDIA CUDA Toolkit 10.2 (for Jetson/AArch64: included with JetPack)
- TensorFlow for Jetson (only on AArch64/Jetson systems, follow the instructions exactly and only install TensorFlow version 2)
- Google Coral AI Toolkit (if you want to use the TensorFlow Lite engine)
- python3-opencv (only on AArch64/Jetson systems)
All the other dependencies are managed by Pip. Once you've satisfied this somewhat onerous list of requirements:
>>> git clone https://github.com/Robocubs/mlrun >>> pip3 install -r requirements.txt >>> python3 -m mlrun [config name here]