mingming715 / deep_features

A ROS package for deep CNN-based feature extraction and publishing

ROS toolkit for deep feature extraction

This repo contains the following ROS packages:

• feature_extraction: real-time extraction of image features (keypoints and their descriptors and scores, and per-image global descriptors)
• image_feature_msgs: definition of feature messages

There are also non-ROS scripts for feature extraction, keypoint visualization and keypoint matching, which do not require ROS installation nor the building procedures in below:

• feature_extraction/show_keypoints.py: show extracted keypoints from given image files
• feature_extraction/show_match.py: show feature matching results from give image pairs

Setup

System requirement

• Ubuntu + ROS
• Python 3.6 or higher
• TensorFlow 1.12 or higher (pip3 install tensorflow)
• OpenVINO 2020 R1 or higher (download)
• OpenCV for Python3 (pip3 install opencv-python; not needed if OpenVINO is installed and activated)
• numpy (pip3 install numpy)
• No GPU requirement

Download and build

0. Preliminary

sudo apt install python3-dev python-catkin-tools python3-catkin-pkg-modules python3-rospkg-modules python3-empy python3-yaml

1. Set up catkin workspace and download this repo

mkdir src && cd src
git clone https://github.com/cedrusx/deep_features_ros.git

2. To use cv_bridge in Python3 with ROS Melodic or older versions, you need to compile it locally. NOT needed if you are using ROS Noetic or newer version.

git clone -b melodic https://github.com/ros-perception/vision_opencv.git

3. Build

# go to your catkin workspace (the parent folder of src)
cd ..
# configure to build for Python3 - Please change the path in the following command according to your Python version
catkin config -DPYTHON_EXECUTABLE=/usr/bin/python3 -DPYTHON_INCLUDE_DIR=/usr/include/python3.6m -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so
# set the parent catkin workspace
. /opt/ros/melodic/setup.bash
# build with catkin from the python-catkin-tools package
catkin build

5. Donwload one of the saved HF-Net models from the Releases, and unzip it.

Run

Feature extraction

Start the feature extraction node, which will subscribe to one or more image topic(s) and publish the extracted image features on corresponding topic(s) with /features suffix.

With OpenVINO model:

. /opt/intel/openvino/bin/setupvars.sh
. YOUR_PATH_TO_CATKIN_WS/devel/setup.bash
rosrun feature_extraction feature_extraction_node.py _topics:=/YOUR_CAMERA_TOPIC _net:=hfnet_vino _model_path:=YOUR_PATH_TO_MODEL_FOLDER

With TensorFlow model:

. YOUR_PATH_TO_CATKIN_WS/devel/setup.bash
rosrun feature_extraction feature_extraction_node.py _topics:=/YOUR_CAMERA_TOPIC _net:=hfnet_tf _model_path:=YOUR_PATH_TO_MODEL_FOLDER

The topics param can take one or more topic names (separated by a comma), e.g. /usb_cam/image_raw, or /left_cam/image_raw,/right_cam/image_raw.

Additional params and their default values (more model-specific params are defined in the default_config dict in the source code):

_keypoint_number:=500 \
_keypoint_threshold:=0.002 \
_gui:=True \
_log_interval:=3.0 \