View Code? Open in Web Editor
NEW
This project forked from shanmo/ros2_tvm
Python 60.79%
Rust 39.21%
ros2_tvm's Introduction
- this repo test the object detection and segmentation models with tvm in ROS2
- ubuntu 20.04, ROS2 galactic
- tvm
- pytorch
- the object detection model is DETR
- I tried many models from mxnet, tensorflow, and pytorch, DETR from pytorch is the only one what works correctly in tvm rust
- the script to generate the tvm graph is in
scripts/test_detection_pytorch.py
- the image segmentation model is unet
- the script to generate the tvm graph is in
scripts/test_seg.py
- generate the models for tvm using the python scripts are in the
scripts
folder
cargo run
- run the ROS2 kitti dataset publisher
- for cpu version
tigerlake
- object detection takes about
478.95 ms
per image, which is 2.06 Hz
- segmentation takes about
356.32 ms
per image, which is 2.81 Hz
ros2_tvm's People
Contributors