Skip to content

knorth55/ros_chainercv

Folders and files

NameName
Last commit message
Last commit date

Latest commit

d4f9688 · Feb 5, 2021
Feb 5, 2021
Nov 14, 2020
Feb 19, 2019
Nov 7, 2020
May 31, 2019
May 31, 2019
Feb 21, 2019
May 31, 2019
Feb 19, 2019
Oct 23, 2018
Sep 4, 2020
Sep 3, 2020
Oct 23, 2018
Nov 7, 2020
Sep 4, 2020
Sep 3, 2020
Sep 3, 2020
Feb 19, 2019

Repository files navigation

ROS-ChainerCV

GitHub version GitHub Workflow Status (branch) Docker Stars Docker Pulls Docker Automated Docker Build Status

This is ROS wrapper of ChainerCV using catkin_virtualenv.

This project depends on locusrobotics/catkin_virtualenv and chainer/chainercv.

Main support

  • Ubuntu 18.04 + ROS Melodic

Docker support

You can get docker images below from Docker hub repository knorth55/ros_chainercv.

Available tags are below.

  • Ubuntu 16.04 + ROS Kinetic
    • Kinetic: kinetic-latest
    • Kinetic + CUDA8.0: kinetic-cuda80-latest
    • Kinetic + CUDA9.0: kinetic-cuda90-latest
    • Kinetic + CUDA9.1: kinetic-cuda91-latest
    • Kinetic + CUDA9.2: kinetic-cuda92-latest
    • Kinetic + CUDA10.0: kinetic-cuda100-latest
  • Ubuntu 18.04 + ROS Melodic
    • Melodic: melodic-latest
    • Melodic + CUDA9.2: melodic-cuda92-latest
    • Melodic + CUDA10.0: melodic-cuda100-latest

You can get docker images as a command below.

docker pull knorth55/ros_chainercv:melodic-cuda92-latest

Build

Build with nvidia-cuda-toolkit deb package

Please build this package using the standard build procedure of ROS.

mkdir catkin_ws/src -p
cd catkin_ws/src
git clone https://github.com/knorth55/ros_chainercv.git
cd ..
rosdep install --ignore-src --from-path src -y -r -i
catkin build
source devel/setup.bash

Build with your own CUDA

mkdir catkin_ws/src -p
cd catkin_ws/src
git clone https://github.com/knorth55/ros_chainercv.git
cd ros_chaienrcv
vim package.xml
# remove nvidia-cuda dependency
vim requirements.txt
# modify cupy-cuda91 to cupy with your CUDA version
cd ../../
rosdep install --ignore-src --from-path src -y -r -i
catkin build
source devel/setup.bash

Run sample

Object Detection

  • FasterRCNN : roslaunch ros_chainercv sample_faster_rcnn_object_detection.launch
  • FasterRCNN FPN: roslaunch ros_chainercv sample_faster_rcnn_fpn_object_detection.launch
  • SSD: roslaunch ros_chainercv sample_ssd_object_detection.launch
  • YOLO: roslaunch ros_chainercv sample_yolo_object_detection.launch

Semantic Segmentation

  • PSPNet: roslaunch ros_chainercv sample_pspnet_semantic_segmentation.launch
  • DeepLab V3+: roslaunch ros_chainercv sample_deeplab_semantic_segmentation.launch

Instance Segmentation

  • FCIS: roslaunch ros_chainercv sample_fcis_instance_segmentation.launch
  • MaskRCNN FPN: roslaunch ros_chainercv sample_mask_rcnn_fpn_instance_segmentation.launch

Tested environment

  • Ubuntu: 18.04
  • ROS: Melodic
  • CUDA: 9.1
  • Chainer: 6.0.0
  • CuPy: 6.0.0
  • ChainerCV: 0.13