Affonso-Gui/detic_ros

Just a simple ROS wrapper for Detic instance segmentation using pre-trained dataset.

detic_ros

ROS package for Detic. Run on both CPU and GPU, GPU is way performant, but work fine also with CPU (take few seconds to process single image).

example of customize vocabulary. Left: default (lvis), Right: custom ('bottle,shoe')

How to running as a node

step1 (build docker container and launch Detic-segmentor node)

Ofcourse you can build this pacakge on your workspace and launch as normal ros package. But for those using CUDA, the following docker based approach might be safer and easy.

Prerequsite: You need to preinstall nvidia-container-toolkit beforehand. see (https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html)

Build docker image

git clone https://github.com/HiroIshida/detic_ros.git
cd detic_ros
docker build -t detic_ros .

Example for running node on pr1040 (please replace pr1040 by you robot hostname or localhost):

docker run --rm --net=host -it --gpus 1 detic_ros:latest \
    /bin/bash -i -c \
    'source ~/.bashrc; \
    rossetip; rossetmaster pr1040; \
    roslaunch detic_ros sample.launch \
    out_debug_img:=true \
    out_debug_segimg:=false \
    compressed:=false \
    device:=auto \
    input_image:=/kinect_head/rgb/image_color'

Change the pr1040 part and /kinect_head/rgb/image_color in command above by your custom host name and an image topic. If compressed image (e.g. /kinect_head/rgb/image_color/compressed) corresponding to the specified input_image is also published, by setting compressed:=true, you can reduce the topic pub-sub latency. device is set to auto by default. But you can specify either from cpu or cuda.

custom vocabulary

Add additional arguments to the script above. example: vocabulary:='custom' custom_vocabulary:='bottle,shoe'.

step2 (Subscribe from node in step1 and do something)

Example for using the published topic from the node above is masked_image_publisher.py. This will be helpful for understanding how to apply SegmentationInfo message to a image. The test file for this example also might be helpful.

ROS node information

• ~input_image (sensor_msgs/Image)
  • Input image
• ~debug_image (sensor_msgs/Image)
  • debug image
• ~debug_segmentation_image (sensor_msgs/Image with 8UC1 encoding)
  • Say detected class number is 14, ~segmentation_image in grayscale image is almost completely dark and not good for debugging. Therefore this topic scale the value to [0 ~ 255] so that grayscale image is human-friendly.
• ~segmentation_info (detic_ros/SegmentationInfo)
  • class name list, confidence score list and segmentation image with 8UC1 encoding. The image is filled by 0 and positive integers indicating segmented object number. These indexes correspond to those of class name list and confidence score list. Note that index 0 is always reserved for 'background' instance and the confidence of the that instance is always 1.0.

As for rosparam, see node_cofig.py.

Running without roscore to batch processing a bag file

rosrun detic_ros batch_processor.py path/to/bagfile

See source code for the options.