Mateusz Płonka

0 %
Mateusz Płonka
Computer Graphics Developer
Motion Designer
  • Residence:
    Poland
  • Region:
    Silesia
  • Age:
    24
Development
Unity
Shaders
Machine Learning
OpenGL
OpenCV
Design
After Effects
Premiere Pro
Photoshop
Blender
  • Computer graphics knowledge
  • C++, C#, Python
  • HLSL, GLSL, GL
  • GIT knowledge
  • Design of graphics and animation

C3D API

Programing

Project details

Description

The c3d-api is dedicated to motion capture data processing, particularly focused on .c3d file interpretation, conversion of data formats, and preparation of datasets for machine learning applications. It features three main Python-based CLI applications, each addressing a specific aspect of motion capture data manipulation.

Components

1. C3D File Interpreter

The C3D File Interpreter is a key component designed to extract and process data from .c3d files used in motion capture systems. This tool focuses on:

  • Data Extraction: Parsing the three segments (header, parameters, and data) of .c3d files to access motion capture data.
  • Selective Data Export: Enabling users to choose specific data sections (markers, virtual markers, modeled markers, angles, forces, moments, and power) for export.
  • CSV Output: Converting and saving the selected data to .csv format for ease of use in applications like Microsoft Excel.
  • Specialized for Joint Angles: Primarily used for exporting joint angles as Cardan angles, facilitating detailed motion analysis.
2. Cardan Angles to Quaternions Converter

This application bridges the gap between traditional angle representations and modern quaternion-based methods used in advanced motion analysis and machine learning:

  • Cardan to Quaternion Conversion: Transforms Cardan angles from .csv files into quaternion sequences, a more robust representation for 3D rotations.
  • Utilization of Scipy Library: Leverages Scipy's scientific computing capabilities for accurate and efficient conversion processes.
  • Structured Output: Produces a .csv file mirroring the structure of the input but with quaternion data, ensuring compatibility with subsequent processing stages.
3. Training Dataset Generator

A crucial tool for preparing machine learning datasets from motion capture data:

  • Dataset Preparation: Generates training and validation datasets tailored for use in recurrent neural network models.
  • Flexible Data Handling: Capable of processing output from both the C3D interpreter and the quaternion converter.
  • Customizable Sequence Length: Allows users to define the sequence length n, essential for configuring datasets for specific machine learning models.
  • Output Files: Creates two .csv files – one for training and another for validation, formatted for direct use in neural network training environments.
  • Start Date:
    July 2023
  • Type
    Toolkit
  • Status:
    Finished
  • Client
    Own project
  • Location:
    Poland, Gliwice

Ready to order your project?

Let's work together!
Contact me