Shared Coordinates

This repository holds a set of tools to help us analyze the data set obtained from the Microsoft Hololens. The aim is to be able to create a shared coordinates systems to optimize data processing and network bandwidth.

How to analyze a data set

You can find an example of a data set at data/process_log. The process to analyze the data set is as follows;

1. Run process_data_set.py to obtain augmented images, marker-based coordinates data, and detection coordinates data.
2. Run process_log.py to obtain raycast-based coordinates data.
3. Run separate_by_position.py to separate images and data from the different positions into different directories.
4. Run calc_stats.py to calculate the statistics of the previously obtained coordinates.

Create Marker

Usage

Command Line

python create_marker.py [-i <id>] [-v] [-s] [-o <output_image>] [-m <size-in-pixels>]

• -i: ID of the marker to be created
• -v: Verbose mode
• -s: Show window with original image augmented with marker coordinates system axis
• -o: Path to where the marker image will be written
• -m: Size of the marker image in pixels

Get Shared Coordinates

A data set of picture for testing this is provided at data/original.

Usage

Command Line

python get_shared_coord.py -i <input_image> -c <camera_config_file> [-v] [-s] [-o <output_image>] [-p <params_file>] [-m <size-in-meters>]

• -i: Path to the input image
• -c: Path to the camera calibration file
• -v: Verbose mode
• -s: Show window with original image augmented with marker coordinates system axis
• -o: Path to where the augmented image will be written
• -p: Path to where the JSON file containing the following information will be written;
  • ID of each of the detected marker
  • Pixel location of the center point of each of the detected marker
  • 3D location of the center point of each of the detected marker
  • Normal vector with respect to each detected marker
• -m: Size of the printed marker in meters

Programmatically

def get_coordinates(in_path,
                    camera,
                    out_path,
                    parameters_path=None,
                    show_window=False,
                    marker_size=0.071):

• in_path: Path to the input image
• camera: Path to the camera calibration file
• show_window: Show window with original image augmented with marker coordinates system axis
• out_path: Path to where the augmented image will be written
• parameters_path: Path to where the JSON file containing the following information will be written;
  • ID of each of the detected marker
  • Pixel location of the center point of each of the detected marker
  • 3D location of the center point of each of the detected marker
  • Normal vector with respect to each detected marker
• marker_size: Size of the printed marker in meters

Circles Distance

A data set of picture for testing this is provided at data/circles_distance.

Usage

Command Line

python circles_distance.py -i <input_image> -c <camera_config_file> [-v] [-s] [-o <output_image>] [-m <size-in-meters>]

• -i: Path to the input image
• -c: Path to the camera calibration file
• -v: Verbose mode
• -s: Show window with original image augmented with marker coordinates system axis
• -o: Path to where the augmented image will be written
• -m: Size of the printed marker in meters

Programmatically

def circles_distance(in_path,
                     out_path,
                     camera,
                     show_window=False,
                     marker_size=0.071):

• in_path: Path to the input image
• camera: Path to the camera calibration file
• show_window: Show window with original image augmented with marker coordinates system axis
• out_path: Path to where the augmented image will be written
• marker_size: Size of the printed marker in meters

Process Data Set

This command assumes that the data set contains multiple images stored in a folder (input path) and with the following files;

• .jpg : Image file
• .json : Server data response
• _answer.json : Detected objects and their center position
• _detection.txt : Bounding box information returned by the Deep Learning algorithm

You can find an example of this data set at data/process_data.

Usage

python process_data_set.py -i <input_path> -c <camera_config_file> [-l <level>] [-m <marker-size-in-meters>]

• -i: Path to the input data set
• -c: Path to the camera calibration file. For the data set at data/process_data, data/calibration/hololens/hololens.yml can be used.
• -l: Logging level possible values are; info, debug, warning and error.
• -m: Size of the printed marker in meters

Process Log

This script processes a log as the one at data/process_data/data.log.

Usage

python process_log.py -i <input_log> [-v] -o <output-path>

• -i: Path to the input log file
• -v: Verbose mode
• -o: Output path to where the obtained information will be placed

Separate By Position

This script separates the images and their corresponding data into different folders. Useful for further analysis. For example, to use the calc_stats.py script.

Usage

python separate_by_position.py -i <input_path> [-v] -s <path-to-sep-file> -o <output-path>

• -i: Path to the input data set
• -v: Verbose mode
• -o: Output path to where the separated data will be placed
• -s: Path to separation file

Calculate Stats

This scripts calculates statistics from the coordinates of a data set.

Usage

python calc_stats.py -i <input_path> [-v] -o <output-path>

• -i: Path to the input data set
• -v: Verbose mode
• -o: Output path to where the statistics output files will be placed