pose_visualizer

pose_visualizer

This project adds a visual layer to the pose_recognition project. More specifically, it displays a 3D hand who can reproduce the different poses being classified by the classification system.

Essentially, this code hooks up to the Redis DB used by the pose_recognition software and renders the recognized hand poses. To do so, this software relies on the libHand library and the OpenCV library for the handling of graphics. Also, note that the cpp_redis library was used to interface with the Redis DB.

Screenshot of the display in action

Requirements

• linux based operating system
• C++ 11
• cmake
• OpenCV
• libHand
• cpp_redis
• pose_recognition

Installing and Running

Installing

# Clone the project and enter the project directory
git clone (URL to the current repo)
cd pose_visualizer
# Create a build directory and move into it
mkdir build && cd build
# Generate the Makefile using CMake and build the executable
cmake...
make

Running

# Launch the executable (from the build folder)
./hand_display

Specifying the hand poses

Before being able to use the executable, the user has to modify the contents of the "Training_data.json" file used by the pose_recognition code. For each specified movement in the info file, the "pose_path" field must specify an absolute path to the corresponding hand pose .yml file.

All the hand pose .yml files currently implemented are located in the "pose_visualizer/hand_poses" directory.

Movement info entry example

{
    "name" : "index_extension",
    "directory_path" : "/home/david/Documents/Synapsets/work_in_progress/Training_data_v2_processed/Index_extension",
    "pose_path" : "/home/david/Documents/GIT/pose_visualizer/hand_poses/index_finger_extended.yml",
    "label" : "0" 
}

Using the executable

Note that the generated executable is not a stand-alone program, it is meant to work alongside the pose_recognition package. The movement classification system should be running before the exec is launched.

Once the exec is running, follow these steps :

• Provide the path to the Training_data.json file provided in the project folder
• Provide the path to the scene specification file (pose_visualizer/hand_model/scene_spec.yml)
• The executable should now be connected to the classification system.