System Introduction

Based on our extensive human movement database, the "Fastmove AI 3D motion analysis system" uses deep learning and neural network algorithm to realize automatic markerless recognition of 21 human body landmarks and a variety of sports instruments. It is a tool to carry out 3D motion analysis, provide the world-leading data analysis content for competitive sports athletes and coaches, and supporting digital solutions for sports science researchers.

Automatic analysis by AI
The "Fastmove AI 3D motion analysis system" provides a more effective option than the traditional manual-labelling video analysis. The time for data analysis can be saved more than sixfold as much. 
Data Accuracy

Red marks are manually labelled landmarks from the referential database. 

White marks are recognized and tracked by the "Fastmove AI 3D
motion analysis system", showing excellent consistency with the reference and providing highly reliable analysis results. 


Software function

Capture videos and clips:

The software supports real-time motion video capture by using ordinary high-speed cameras or industrial cameras. Multi-camera synchronization, clipping, and exporting functions are also provided.

Spatial calibration:

The software provides Direct Linear Transformation (DLT) calibration and other calibration schemes, making 3D spatial calibration and coordinate system transformation possible. 


3D data synthesis:

The software can synthesize 3D coordinates of human body landmarks by aligning synchronized images captured by two or more cameras. Integrated with spatial calibration and 3D motion tracking, the software can provide 3D coordinates data and help athletes and coaches comprehensively analysis movement and improve performance.

Manual correction:

The software can manually correct the errors, confusions and incorrect points in automatic recognition through manual correction function. It provides a variety of convenient manual correction tools including two-dimensional coordinate correction, automatic removal of error recognition, left-right identification exchange, overall smooth filtering, etc.


Manual Correction Full Demo

After AI automatic recognition, manual correction is carried out by using tools such as coordinate correction, automatic removal, smoothing filtering, etc., which makes the data more refined and significantly saves analysis time
Four-shot matrix
———————————

The "Four-shot matrix" module uses four cameras to create an omnidirectional motion capture volume. It can effectively reduce the inaccurate recognitions caused by occlusions and further improve the efficiency of data processing. 


Ultra-complex scene synthesis solution
Project case: steel frame snowmobile
——————————————————————
We use up to six high-speed cameras to perform 3D reconstruction of ultra-complex scenes (height drop, long and narrow distance), and to reconstruct three-dimensional data using AI technology

Hardware configuration
Software configuration