Sadaku1993 / deeplab_ros

Deeplab for ROS

In this repository, I compiled the source code using ROS and Pyhon2.7

Overview

DeepLab: Deep Labelling for Semantic Image Segmentation for ROS

Requirements

• ROS Kinetic(ubuntu 16.04)
• Python2.7+
• Opencv3.3+
• tensorflow1.4+
• DeepLabv3

How to Install

DeepLab: Deep Labelling for Semantic Image Segmentation

• DeepLab

ROS Kinetic

• ROS Kinetic installation

Clone this Repository

$ git clone https://github.com/Sadaku1993/deeplab_ros
$ cd catkin_ws
$ catkin make

Download model

See the TensorFlow DeepLab Model Zoo

$ roscd deeplab_ros
$ cd deeplab
$ wget http://download.tensorflow.org/models/deeplabv3_mnv2_pascal_train_aug_2018_01_29.tar.gz 

How to RUN

Change model name

$ roscd deeplab_ros/deeplab
$ vim deeplab_ros.py

deeplab_ros.py

    MODEL_NAME = 'deeplabv3_mnv2_pascal_train_aug_2018_01_29.tar.gz'
    # MODEL_NAME = 'deeplabv3_mnv2_cityscapes_train_2018_02_05.tar.gz'

Change param

$ roscd deeplab_ros/launch
$ vim deeplab_ros.launch

deeplab_ros.launch

image : Subscribe Topic(sensor_msgs/Image)
deeplab/image : Publish Topic(sensor_msgs/Image)

Launch rosnode

$ roslaunch deeplab_ros deeplab_ros.launch