STREACKER Project (Top)
STREACKER Project (Skeletal TRacking Enhanced with Anatomically Corrected Kinematics for Exergames and Rehabilitation) aims the development of a new
algorithm for skeletal tracking with anatomically correct segment orientation based on the combination of vector orthogonalization and advanced machine learning techniques.
The final objective is to augment the use of current optical markerless motion capture systems (OML-MOCAPs) in the field of physical rehabilitation, considering an interactive
platform which would assist the therapists during the rehabilitation sessions (e.g. motion analysis, prescribe exercises, exergames).
The project results from the collaboration of several researchers from INESC-ID, IDMEC, Instituto Superior Técnico (Universidade de Lisboa) and the University
of Texas at Austin.
For more information regarding the STREACKER Project, please visit https://streacker.inesc-id.pt
or consult the STREACKER reference paper: https://doi.org/10.1016/j.patrec.2022.12.012
The STREACKER source code can be downloaded at:
- Google Colab
- GitHub
Selecção de Movimentos (Movements Selection) (Top)
In the scope of the STREACKER project, a set of exercises were experimental acquired using an Optoelectronic Motion Capture System (Qualisys ProReflex 1000) to be used during the training and testing of the developed algorithms.
The list of movements to be acquired was achieved by applying the selection methodology described in supplementary materials [1]. In summary, an initial list of approximately 100 movements performed in fitness and gym workouts was created
by a professional fitness trainer. After a first selection procedure, based on the inclusion and exclusion criteria defined for the study, the list was reduced to approximately 50 movements.
The movements were posteriorly rated by 10 fitness trainers with more than 5 years of expertise in the area using a Likert scale (1 to 5). The movements with a rating higher than 4 advanced for the next selection step.
A second removal procedure was done by excluding movements that could result in experimental problems during the data acquisition (occlusion or destruction of markers, occlusion of body segments, etc,).
In the end, 10 movements were selected and divided in three major groups: lunges & variants, squats & variants and other movements (see list below).
Two more sequences of multiplanar movements were posteriorly included in the list for training and testing of the developed methodologies.
[1] Fernandes, F., Roupa, I., Gonçalves, S. B., Moita, G., da Silva, M. T., Pereira, J., Jorge, J., Neptune, R. R., & Lopes, D. S. (2023).
Sticks and STONES may build my bones: Deep learning reconstruction of limb rotations in stick figures. Pattern Recognition Letters, 165, 138–145.
https://doi.org/10.1016/j.patrec.2022.12.012
Supplementary Materials (description of the procedure for the selection of the movements list): https://www.sciencedirect.com/science/article/pii/S0167865522003804?via%3Dihub#ecom0001
Lista de Movimentos (Movements List) (Top)

Aquisição de Movimentos (Movement Acquisition) (Top)

Acquisition Information |
|
Local: | LBL - Laboratório de Biomecânica de Lisboa |
System: | Qualisys ProReflex 1000 |
Nr. of Cameras: | 14 |
Acquisition Frequency: | 100 |
Number of markers: | 68 |
Acquisition Software: | Qualisys Track Manager 2.9 (build 1563) |
Ethics Approval: | The study was submitted to the ethics committee of Instituto Superior Técnico and was approved in January 2020 (Ref. nr. 1/2020 (CEIST)) |
Volunteers Information |
|||
Male | Female | Total | |
Number of Volunteers: | 9 | 7 | 16 |
Age: | 29.6±11.3 | 24.6±5.7 | 27.4± |
Height (m): | 1.78±0.07 | 1.60±0.08 | 1.7± |
Weight (kg): | 70.6±9.7 | 55.1±6.9 | 63.8± |
IMC: | 22.2±2.2 | 21.4±1.2 | 21.9± |
Fitness/Exercise Professionals (%): | 11± | 29± | 18.9± |
Practice Physical Exercise (>2 days/week) (%) | 33± | 57± | 43.5± |
Each volunteer performed at least 7 repetitions of each movement. For unilateral movements, the movement was acquired for both sides (e.g.. right and left lunge, right and left curtsy lunge etc.).
Considering all the repetitions from the 16 volunteers, a total of 761 234 frames were acquired and included in the STREACKER database.
Ficheiros da Base de Dados STREACKER (STREACKER Database Files) (Top)
Information available on database files:
- Volunteer Information
- Coordinates of the Joints and Extremities
- Reference Frame of Segments - Rotatation Matrix (Order: M11 M12 M13 M21 M22 M23 M31 M32 M33)
- Reference Frame of Segments - Euler Angles (Order: z y x (in deg))
- Relative Rotation Matrix (Parent->Child) - Rotation Matrix (Order: M11 M12 M13 M21 M22 M23 M31 M32 M33)
- Relative Euler Angles (Parent->Child) - Rotation Matrix (Order: z y x (in deg))
- Joint Angles (Order: x y z (in Deg))
- Experimental data was smoothed using a 2nd order Butterworth filter with a cutoff frequency of 6 Hz
List Of Points (Joints and Extremities) | |||
1 - Right Toes | 7 - Left Toes | 13 - Spine Base (Mean HJ) | 19 - Right Wrist |
2 - Right Meta | 8 - Left Meta | 14 - Mid Spine | 20 - Right Hand |
3 - Right Heel | 9 - Left Heel | 15 - Neck | 21 - Left Shoulder |
4 - Right Ankle | 10 - Left Ankle | 16 - Top Head | 22 - Left Elbow |
5 - Right Knee | 11 - Left Knee | 17 - Right Shoulder | 23 - Left Wrist |
6 - Right Hip | 12 - Left Hip | 18 - Right Elbow | 24 - Right Hand |
List Of Segments | |||
1 - Right Toe | 6 - Left Hindfoot | 11 - Upper Trunk | 16 - Right Forearm |
2 - Right Hindfoot | 7 - Left Shank | 12 - Neck & Head | 17 - Right Hand |
3 - Right Shank | 8 - Left Tigh | 13 - Right Scapula/Clavicule | 18 - Left Upperarm |
4 - Right Tigh | 9 - Pelvis | 14 - Left Scapula/Clavicule | 19 - Left Forearm |
5 - Left Toe | 10 - Mid Trunk | 15 - Right Upperarm | 20 - Left Hand |

Transferência da Base de Dados (Database Download) (Top)
If you intended to download the STREACKER database, please, fulfill the following form. The link for the download of the database files will be sent for the submitted email.
Please, cite the following papers, if you intend to use the STREACKER database for R&D purposes: Fernandes, F., Roupa, I., Gonçalves, S. B., Moita, G., da Silva, M. T., Pereira, J., Jorge, J., Neptune, R. R., & Lopes, D. S. (2023). Sticks and STONES may build my bones: Deep learning reconstruction of limb rotations in stick figures. Pattern Recognition Letters, 165, 138–145. https://doi.org/10.1016/j.patrec.2022.12.012