Enhancing Speed Climbing Performance and Optimizing Training

Methods Through Advanced Video Analysis

Dominik Pandurevic

1 a

, Pawel Draga

1 b

, Klaus Hochradel

1 c

, Lewis Chew

, Muhammad Hidayat

and Alexander Sutor

1 d

Institute of Measurement and Sensor Technology,

UMIT TIROL - Private University for Health Sciences and Health Technology, Hall in Tirol, Austria

National Youth Sports Institute, Singapore

Keywords:

Sports Technology, Sports Science, Speed Climbing, Performance Analysis, Human Pose and Feature

Detection, Training Optimization.

Abstract:

This paper presents the implementation of an automated recording and analysis system for the capturing and

evaluation of performances of speed climbing athletes. In collaboration with the National Youth Sports Insti-

tute (NYSI) in Singapore, an advanced camera system was installed on a newly constructed speed climbing

wall, aiming to enhance training protocols by providing detailed performance insights. The paper is mainly

split into two parts: the ﬁrst one describes the hardware setup, including camera selection, conﬁguration, inte-

gration with the existing infrastructure and data collection methods. The second part presents the analysis of

performance data of one athlete trained at the NYSI, highlighting key ﬁndings and potential training improve-

ments. Preliminary results indicate signiﬁcant beneﬁts in technique reﬁnement and performance optimization,

demonstrating the system’s value in a competitive training environment. Through section wise analysis of

several runs of the same athlete, a stagnation in the start section of the standardised wall has been detected.

However, a signiﬁcant improvement was determined in the other sections, which led to a noticeable overall

increase in performance in less than 2 years of regular training.

1 INTRODUCTION

Climbing as a sport is experiencing rapid growth. The

International Federation of Sport Climbing (IFSC) re-

ports that approximately 25 million people of all ages

now climb regularly around the world. Between 2001

and 2012, the global number of climbers and climb-

ing facilities surged by nearly 50% (Grønhaug and

Norberg, 2016). The sport has also gained greater

recognition in the media, and the number of newly

built climbing centers continues to rise, reﬂecting

its increasing popularity. The acknowledgment on

an international level is also remarkable, especially

with its debut as an Olympic discipline at the 2020

Tokyo Olympics and its continued presence in the re-

cently concluded 2024 Olympics in Paris. Whereas

in Tokyo all three disciplines Bouldering, Lead and

https://orcid.org/0000-0002-7801-2090

https://orcid.org/0000-0002-1359-6806

https://orcid.org/0000-0003-4695-5809

https://orcid.org/0000-0003-0064-5493

Speed Climbing were combined within one compe-

tition, this mode has changed in Paris with the sep-

aration of speed climbing as an independent disci-

pline. These disciplines differ in terms of execution

and loading, as highlighted by (Winkler et al., 2022),

as well as the power demands of the upper limbs, em-

phasized by (Levernier et al., 2020). Despite these

differences, the results in Speed Climbing for both

men and women in Paris showed signiﬁcant improve-

ments compared to Tokyo. For example, the previ-

ous Olympic record improved from 6.97s to 6.06s for

women and from 5.45s to 4.74s for men (International

Federation of Sport Climbing, 2024). This success

was unexpected, considering the stagnation of top-

end times in the years following Tokyo. This break-

through shifted the focus to reﬁning technique and op-

timizing performance, especially for young athletes

aiming to compete at the highest levels.

Speed Climbing continues to evolve in various

ways: athletes are shortening the climbing route by

skipping certain holds (Pandurevi

c et al., 2022) and

optimizing the starting position to enhance their efﬁ-

Pandurevic, D., Draga, P., Hochradel, K., Chew, L., Hidayat, M. and Sutor, A.

Enhancing Speed Climbing Performance and Optimizing Training Methods Through Advanced Video Analysis.

DOI: 10.5220/0013060000003828

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 12th International Conference on Sport Sciences Research and Technology Support (icSPORTS 2024), pages 233-240

ISBN: 978-989-758-719-1; ISSN: 2184-3201

233

ciency. Similar adjustments are also being made in

the ﬁnal phase of the climb, which helps reduce time

and improve overall performance. These innovations

are driving the dynamic progression of the discipline.

As Speed Climbing advances, the technology used to

train the next generation of athletes must keep pace

with these changes.

This study is based on the collaboration with the

National Youth Sports Institute (NYSI) in Singapore,

an institution dedicated to the development of youth

athletes across various sport disciplines. Utilization

of NYSI’s modern facilities including a newly es-

tablished standardized speed climbing wall, an au-

tomated camera system was designed and imple-

mented to provide detailed insights into climbing per-

formance. Accordingly, this paper is structured in two

parts, namely

• the design and implementation of a robust cam-

era system capable of recording speed climbing

videos with little effort and

• the evaluation of recorded data to identify patterns

and insights for the enhancement of training meth-

ods for speed climbing athletes.

The hardware was designed with mobility and sim-

plicity in mind, so that the system can be used ﬂexibly

depending on the application. Since the athletes and

trainers should concentrate on the sport itself, the sys-

tem is intended to reduce the effort involved in record-

ing and subsequent video analysis.

There are already several research groups work-

ing on similar topics related to the performance anal-

ysis in sport climbing. Besides several approaches

focusing on the usage of embedded climbing holds

with a wide variety of instrumentation solutions, there

are also some methods presented on how to track the

body of sport climbers in different ways. Most com-

mon is the application of markers on several positions

on the body (Reveret et al., 2020). Using a drone-

based system tracking a marker placed on the harness

of the speed climbing athlete, this approach enables

the determination of the body’s 3D position and ve-

locity. Compared to that, (Beltr

an et al., 2022) pre-

sented a method for the segmentation of human body

movements in sport climbing using kinematic vari-

ables obtained from RGB-D cameras, combining the

rendered RGB images using different pose estimators

with the aligned depth frames. Legreneur et al., as

one of the ﬁrst, (Legreneur et al., 2019) manually an-

alyzed speed climbing runs of athletes participating

at the Youth Olympics 2018 using the open-source

video analysis tool Kinovea. This approach was fur-

ther developed in (Elias et al., 2021) creating a data

set of 362 speed climbing runs including 55 world

elite climbers with the help of the human pose esti-

mator OpenPose (Cao et al., 2018).

Similarly automated movement analysis are also

carried out in other sports. (Evans et al.,

2021) presents two methods for precisely measuring

sprinter’s foot-ground contact locations and timing

using multi-camera systems, where one of the pre-

sented approaches relies on machine learning-based

human pose estimation using OpenPose.

This research aims to contribute to the ﬁeld of

sport science by demonstrating the practical appli-

cation of video analysis technology in speed climb-

ing. It should offer valuable insights into perfor-

mance analysis and possibilities for the optimization

of training methods to keep up with the rapid growth

of this Olympic discipline. The surprising surge in

performance observed at the Olympic Games in Paris

2024 highlights the importance of continuous investi-

gation and improvement of existing training methods

and the replacement of commonly used manual and

time-consuming video analysis with automated sys-

tems based on neural networks. The presented results

reﬂect the need for such systems and demonstrate the

relevance of this research.

2 METHODOLOGY

As already mentioned, this paper is ﬁrstly based on

the design and implementation of a hardware sys-

tem for the automated recording and analysis of speed

climbing videos.

Figure 1 roughly summarizes the implemented

camera system consisting of a remote-controlled

recording unit and a synchronized user interface (UI)

running on the related computer with live preview of

the camera output and several functionalities for the

adjustment of the scene. Apart from starting, stop-

ping and initiating the upload to a cloud via remote or

UI control, the workﬂow from recording to receiving

extensive data packages is fully automated.

2.1 Camera System Design

2.1.1 Housing

In order to accommodate the entire measurement sys-

tem consisting of computer, camera and power sup-

ply units in one compact housing, a chassis with the

dimensions 239 x 300 x 160 mm with a transparent

lid was chosen. Particular emphasis was placed on

the usage in outdoor areas with a sealed closure ca-

pability. Due to the speciﬁc climate in Singapore and

the associated high humidity, the entry of humid air

icSPORTS 2024 - 12th International Conference on Sport Sciences Research and Technology Support

234

(a) (b)

Figure 1: Illustration of the camera system with (a) Side

view of the speed climbing wall and visualization of the

remote-controlled camera setup and (b) the programmed

user interface with snapshot of the camera recording with

enabled focus peaking function.

into the interior of the housing must be prevented, as

this could cause damage to the built-in electronics.

In order to be able to cool the heating measurement

system properly, a fan was installed on the underside,

whereby the air can circulate from the inside to out-

side through two holes covered by ventilation grilles.

2.1.2 Processor

For controlling the camera, receiving and process-

ing signals from the remote control and uploading

recorded speed climbing runs to the conﬁgured cloud,

an Intel i5 NUC 11 Pro Kit with 16 Gb memory is

installed. Its compact size of 117 x 112 x 37 mm en-

sures an easy installation in the housing with com-

paratively high performance output. An additional

8” touchscreen monitor from Magedok has been in-

stalled for troubleshooting and obtaining feedback.

Moreover, it allows manual operation of the pro-

grammed UI on a 720p display if the remote control

should fail.

2.1.3 Sensor and Lens

The choice of an appropriate camera was again

based on both compact size as well as perfor-

mance/usability. In order to obtain high-quality re-

sults from the backend system, at least HD quality

(1080p) and 30 FPS are required for recordings in

vertical format (recording of the entire speed climb-

ing wall, see Figure 1b). The industrial USB camera

Alvium 1800 U-511C manufactured by Allied Vision

fulﬁlls all these minimum requirements and, with a

size of 29 x 29 x 38 mm, ﬁts well into the overall

system. The compatible LMVZ4411 lens from Kowa

completes the camera unit offering manual focus and

zoom adjustment with a focal length range of 4.4 - 11

mm.

2.1.4 Remote Control

The camera setup is completed with the implementa-

tion of a wireless remote system for controlling the

recording and upload processes. To protect the built-

in electronics from impact and aforementioned hu-

midity, a sealed hand-held housing (165 x 90 x 34

mm) has been selected. It can store both an ESP32

Feather V2 Adafruit microcontroller with associated

circuit board and a LiPo battery pack with 6600 mA h

capacity. In addition, 3 push buttons with built-in led

indicators for start, stop and upload functionality as

well as a led light for monitoring the battery charge

status were installed on the lid of the housing. The

remote control was designed to enhance the usability

of the system during training sessions, minimizing the

need for direct interaction with the camera hardware.

2.1.5 User Interface

The programmed UI provides direct control over the

camera with buttons for starting, stopping and upload-

ing, enabling operations without the need for the re-

mote control. It features a live preview of the cam-

era output and is synchronized with remote to ensure

that triggered actions are mirrored in the interface. In

addition, the interface includes modiﬁers like sliders

and buttons for the adjustment of exposure and gain

parameters of the camera (see Figure 1b). Users can

disable the, by default set, continuous automatic ad-

justment by the camera, or manually set these param-

eters with a deﬁned range for precise control over the

recording settings. To adjust the focus of the cam-

era using the mechanical controls on the camera, a

focus peaking function can be enabled, which high-

lights sharp edges with red lines on the recording and

thus supports this manual adjustment.

2.2 Integrated Control System

This subsection introduced the developed control sys-

tem to manage the recording and data management

processes. Thereby, the system consists of both the

mentioned wireless remote control powered by an Ar-

duino device and a Python-based UI for manual oper-

ation and troubleshooting.

Enhancing Speed Climbing Performance and Optimizing Training Methods Through Advanced Video Analysis

235

2.2.1 Arduino-Controlled Remote System

The Arduino microcontroller forms the core of the re-

mote system and allows users to start, stop and up-

load recordings remotely. This system uses WiFi

communication to interact with the camera setup,

whereby the microcontroller initially connects to the

computer via provided mobile hotspot. This allows

the recording process to be both efﬁcient and user-

friendly. The data transfer is realized using MQTT

(Message Queuing Telemetry Transport). MQTT is a

lightweight messaging protocol based on a publish-

subscribe approach. Its application area designed

for low-bandwidth, high-latency networks is ideally

suited for this control system. Devices (clients) pub-

lish messages to deﬁned topics on a broker (server),

which then distributes theses messages to all sub-

scribed clients, enabling efﬁcient, real-time data ex-

change.

The main tasks of the microcontroller is the estab-

lishment of a connection with the computer and the

management of publish and subscribe topics for the

processing of trigger signals of the push buttons. In

the Arduino code, actions for sending messages for

start, stop or upload are triggered with related topics

when the corresponding push buttons are pressed. For

manual execution via UI, subscribe topics are deﬁned

to synchronize the Arduino and the computer, ensur-

ing both devices stay in sync for the same actions. In

addition, a publish topic is speciﬁed, deﬁning the bat-

tery charge status of the installed LiPo.

2.2.2 Signal Processing and Direct Camera

Control

On the computer of the camera system, a python

script is executed, which mainly handles receiving

and processing signals from the remote control, man-

aging the camera through Vimba library, capturing

frames and processing them into a suitable video. Ad-

ditionally, the script runs the mentioned UI, set up for

direct camera control in case of remote control failure

or for camera setup testing. A focus peaking function

has also been implemented to support the manual ad-

justment of the lens using the mechanical controls on

the camera (zoom and focus) setting optimal focus.

This function highlights areas of the image that are

in sharp focus by detecting edges via Canny detec-

tor and overlaying a red color ﬁlter (see Figure 1b).

Therefore, the focus is optimally adjusted when the

whole image or areas of interest have the most dis-

tinctive and extensive highlights.

As already mentioned, the Arduino and Python

components are both designed to work in tandem,

providing a robust and ﬂexible control system for the

recording of speed climbing recordings. While the

Arduino programmed remote offers fast and effortless

operation provided primarily for trainers, the Python

UI enables precise adjustment of camera settings and

diagnostic tasks can be handled effectively. This dual-

system approach increases the reliability and versatil-

ity of the control unit and ensures continuous opera-

tion during speed climbing runs.

(a) (b)

Figure 2: Summary of the image processing from captur-

ing athletes on both sides of the speed climbing wall to

cropped images for further processing; (a) Overlaid images

of two athletes at different heights on the speed climbing

wall tracked with MMDet (Chen et al., 2019); (b) Resulting

crop of the recordings to isolate both athletes for the subse-

quent human pose estimation with MMPose (Contributors,

2020).

2.3 Backend System for Human Pose

Estimation and Feature Detection

Once a video of a speed climbing run has been

recorded with the described camera system, this out-

put requires post-processing to enable further anal-

ysis. In order to achieve equivalent results in each

area of the climbing wall regardless of the height, the

recording is cropped such that the respective athlete is

focused on both sides (see Figure 2b). To achieve this,

the application MMDetection (Chen et al., 2019) is

required ﬁrst. This powerful framework is commonly

used to detect humans as objects in images by apply-

icSPORTS 2024 - 12th International Conference on Sport Sciences Research and Technology Support

236

ing models pre-trained on large datasets like COCO

(Lin et al., 2015). By the detecting the human body

of both athletes and marking them with surrounding

rectangles (see Figure 2a), the resulting cropped im-

ages are a result of moving a window over the whole

route map aligned at the height of the rectangles’ cen-

ter. Furthermore, the two sides of the speed climbing

wall are split by detecting climbing holds and deter-

mining the dividing line. By using the object detec-

tor YOLO and a self-trained model, the holds placed

furthest to the left and right are ﬁrst determined in or-

der to calculate the edges as well as the center line

of the entire speed climbing wall. By combing the

cropped images, a video is created which follows the

athlete for each side similar to camera-guided record-

ings. This inevitably leads to a reduction of the res-

olution from 1920x1080 px to 720x1280 px for each

frame, whereby the remaining width of 200 px is sup-

plemented by adding black bars on both sides of each

image.

By using the top-down approach for human pose

estimation consisting of MMDetection and MMPose

(Contributors, 2020) as well as other feature detectors

described in (Pandurevi

c et al., 2022), movement re-

lated parameters may be calculated. The previous step

eliminates thereby the need for renewed detection hu-

man body surrounding rectangles using MMDetec-

tion. In addition, the calculation step of detecting and

matching random feature points in each image to es-

timate the absolute camera movement is skipped, as

this vector has already been determined for both sides

by the deﬁned window movement within the cropping

process.

The described backend processes run on a external

server with a powerful processor and graphics card to

ensure rapid calculations of the data for an immediate

feedback for the trainers and athletes. Finally, the an-

alyzed data packages are transferred to the same cloud

system used for uploading the recordings, which can

then be evaluated using the provided analysis soft-

ware.

2.4 Analysis Tool

Along with the designed camera system, the NYSI

was provided with an analysis tool for the visualiza-

tion of the calculated data sets (see Figure 3). Due to

the extensive range of computable parameters by the

backend system, the data sets are broken down to the

most critical metrics. The software enables the com-

parison of joint angles and velocities, as well as po-

sition, velocity and acceleration of the center of grav-

ity (COG). For better understanding and correlation

with the data, each data point can be linked to the cor-

responding image of the athlete with overlaid body

skeleton. In addition to the mentioned data plots, a

visualization of the route map with the path of the

COG is also provided. By clicking on the visualized

climbing holds, the jumping to speciﬁc positions in

the data set is enabled, allowing the synchronization

of athletes at different heights for a detailed analysis

of certain sections on the speed climbing wall. The

software allows the comparison of up to two athletes

simultaneously from one or different recordings.

Figure 3: Representation of the analysis tool with visualiza-

tion of the data sets of two athletes from different record-

ings.

3 RESULTS

To validate the above described measurement system,

recorded data sets over 1.5 years were analyzed. Var-

ious parameters were extracted and the progression

of the performance was evaluated. Table 1 displays

an outtake of the statistical analysis of several biome-

chanical parameters of a speed climbing athlete, with

data collected at the NYSI in two different periods:

March 2023 and July 2024. The 23 year old athlete

has been trained at the NYSI since his youth and con-

tinues to do so due to his potential. The analysis fo-

cuses on various performance indicators, namely end

time, velocity, path length, limb frequencies and con-

tact times. Due to their importance and inﬂuence, the

velocity and path data were considered in even further

detail by additionally evaluating the individual sec-

tions of the speed climbing wall (start, middle, end).

Thereby, the start section covering the ﬁrst 5 holds of

the speed climbing wall, the middle section holds 6

to 12 and the end section the remaining ones. For the

frequencies and contact times, the values for hands

and feet have been combined.

Starting with the most informative parameter, the

end time ranged at the beginning of our measurement

time period in March 2023 from a minimum of 6.77

s to a maximum of 8.53 s, with an average of 7.32 s.

Compared to that, by July 2024, the end time showed

Enhancing Speed Climbing Performance and Optimizing Training Methods Through Advanced Video Analysis

237

a signiﬁcant improvement, varying around an average

of 5.99s, reﬂecting a reduction of 18.09 %, indicating

a remarkable increase in performance.

The velocity experienced a slight decrease in the

starting section of the route map, dropping from an

average of 2.19 m/s in March 2023 to 2.18 m/s. How-

ever, this parameter increased signiﬁcantly in the mid-

dle and end section and thus also in the entire area

of the wall. In terms of numbers, the velocity in-

creased by an average of 18.51 % in the middle sec-

tion, 41.77 % in the end section and 25.13 % over the

entire movement process.

(a) (b) (c)

Figure 4: Comparison of the route map (position and veloc-

ity of the COG) of the same athlete with section-wise and

overall path lengths measured in March 2023 (a) and July

2024 (b); (c) shows the overlay of both route maps with blue

circles marking points of signiﬁcant change in technique.

Figure 4a and 4b each show one run from the data

sets measured in March 2023 and July 2024. De-

spite relatively small reduction in path length in the

start and middle sections (see Table 1), Figure 4c in-

dicates an obvious improvement in motion sequences

in all wall sections. However, particularly striking in

the statistical analysis is the almost unchanging over-

all path length by 0.03 %. Since the calculated path

length in the end area includes the positions up to the

top buzzer, the athlete was aiming for different end

positions on these two days. When comparing the

video footage, it is noticeable that on training days

in March 2023 no end buzzer was mounted and the

athlete’s last jump thus did not turn out high enough.

The analysis of the limb frequencies reveals both

an increase and a decrease. While the frequency of

the feet reﬂects an improvement of 12.05 % and thus a

signiﬁcant increase in performance, the average value

of the hands falls from 1.84 Hz to 1.56 Hz and in rel-

ative terms by 15.50 %. This indicates a shift in the

athlete’s technique, whereby the movement of the feet

became faster, while the hands slowed down conspic-

uously.

Finally, it should be emphasized that the contact

times between hands and feet also varied consider-

ably. There was a signiﬁcant decrease of 21.21 % for

the hands and 13.58 % for the feet, indicating a more

efﬁcient and faster climbing technique.

4 DISCUSSION

The analysis of the speed climbing athlete’s per-

formance, including the data from Table 1 between

March 2023 and July 2024 reveals several critical in-

sights into his performance. The individual results of

the measured parameters are discussed in more detail

in the following subsections. Lastly, the limitation of

the measurement system are described.

4.1 Velocity

The almost constant velocity with a slight decrease

recorded at the end of the measurement period (-0.56

%) may not be indicative of a true performance de-

cline. This change could be attributed to daily ﬂuc-

tuations in the athlete’s physical state. This stagna-

tion indicates that the training focus was placed on

other wall sections and no further improvement of the

start technique was achieved. However, the athlete

was able to compensate this with signiﬁcant increases

in the middle and end section, where velocity jumped

by 18.51 % and 41.77 %, respectively. The overall

improvement from 1.63 m/s to 2.04 m/s, refers to en-

hanced efﬁciency and power as the athletes progresses

on the wall.

4.2 Path Length

The slight increase in path length, particularly no-

ticeable in the end section, is mainly attributed to a

icSPORTS 2024 - 12th International Conference on Sport Sciences Research and Technology Support

238

Table 1: Summary of the descriptive statistical evaluation of a NYSI athlete in the period from March 2023 to July 2024. In

addition to the values for minimum (Min), maximum (Max), median (Avg) and standard deviation (Std), the relative change

of the average values in % is also provided.

Time in s Velocity in m/s Path Length in m Frequency in Hz Contacts in s

Statistics End Time Start Middle End Total Start Middle End Total Hands Feet Hands Feet

March 2023

Min 6.77 2.11 1.81 1.20 1.53 3.44 4.42 4.39 12.45 1.70 0.93 0.52 0.43

Max 8.53 2.27 2.03 1.70 1.77 3.83 4.56 4.54 12.80 2.13 1.05 0.77 0.53

Avg 7.32 2.19 1.93 1.49 1.63 3.74 4.51 4.48 12.71 1.84 0.98 0.55 0.44

Std 0.49 0.05 0.06 0.15 0.09 0.12 0.04 0.05 0.10 0.13 0.04 0.09 0.04

July 2024

Min 5.97 2.16 2.25 2.10 1.98 3.70 4.39 4.48 12.70 1.50 1.10 0.40 0.38

Max 6.02 2.20 2.33 2.11 2.10 3.72 4.54 4.59 12.72 1.61 1.11 0.47 0.39

Avg 5.99 2.18 2.29 2.11 2.04 3.71 4.47 4.53 12.71 1.56 1.10 0.43 0.38

Std 0.03 0.02 0.04 0.01 0.06 0.01 0.07 0.07 0.05 0.05 0.01 0.03 0.01

Relative Change in % -18.09 -0.56 18.51 41.77 25.13 -0.85 -0.83 1.22 0.03 -15.50 12.05 -21.21 -13.58

change in sensor placement in training to comply with

IFSC regulations. This adjustment likely extended the

measured path length at the end of the runs recorded

in July 2024. While the increase in path could theoret-

ically challenge the athlete’s velocity, the simultane-

ous reduction of the end time suggests that the athlete

has adapted to this change effectively. Nevertheless,

future training sessions should be standardized to en-

sure data comparability.

4.3 Limb Frequency and Contact Time

Thereby, the most pronounced improvement was

measured by the limb frequency of the feet, which in-

creased by 12.05 %. This enhancement is crucial for

the reduction of contact times, a key factor not only

in speed climbing but also in sprinting. The decrease

of frequency of the hands could indicate a shift to-

wards more intentional and efﬁcient hand placement,

reducing unnecessary movements and optimizing the

athlete’s run.

Getting back to sprinting, the ground contact time

(GCT) is a critical factor for the evaluation of per-

formance. Studies by Hunter et al. (Hunter et al.,

2004) and Mero et al. (Mero et al., 1992) reveal that

elite sprinters have shorter GCTs, enabling maximal

force generation and rapid transitions between single

steps. Similarly in speed climbing, the reduced con-

tact times for both hands (-21.21 %) and feet (-13.58

%) indicate less time in contact with the climbing

wall, which in turn is directly related to an increase

in velocity.

The concept of vertical ground reaction force

(GRF) (Hunter et al., 2005) is likewise relevant.

While faster sprinters generate a comparable vertical

GRF to slower sprinters, they achieve this in a shorter

GCT (Weyand et al., 2000), resulting in a longer step

length and a higher horizontal velocity. The athlete’s

increased feet frequency and reduced contact times

mirror this principle, pointing out a more efﬁcient

transfer of force and faster transition between move-

ments.

4.4 System Limitation

Although the measurement system was designed to be

mostly automated, manual readjustments of the cam-

era must be carried out before each measurement ses-

sion. Therefore, the camera has to be set using the

two parameters exposure and gain, and the focus of

the lens needs to be adjusted. Furthermore, in the

future, an improvement of the camera could aim for

higher resolutions and frame rates. Currently, record-

ings are limited to a maximum of 1080p in order to

maintain 30 FPS. After cropping the single frames as

explained in Figure 2, only a maximum of 640 x 720

px remains to detect each athlete and various features

in the image.

5 CONCLUSION

The introduced measurement system enables trainers

and athletes at the NYSI to monitor relevant move-

ment parameters and their evaluation in single wall

sections to optimize individual training methods. The

user-friendly handling of the wireless camera system

ensures the recording and calculation of the data in

real time and its visualization using the provided anal-

ysis software. By using this system since its intro-

Enhancing Speed Climbing Performance and Optimizing Training Methods Through Advanced Video Analysis

239

duction in March 2023, several data packages from

different athletes have already been calculated. The

results of one of these promising athletes still train-

ing at the NYSI were presented in this study. The

improvements observed, especially in the middle and

end section of the speed climbing wall, highlight the

athlete’s positive adaptation to training methods and

enhanced techniques. The conspicuous reduction in

end time and improvement in most of the presented

parameters conﬁrms this overall improvement from

March 2023 to July 2024. Nonetheless, motion se-

quences, especially in the start section should be ex-

amined in more detail and training methods should be

adapted accordingly in order to achieve an additional

increase in performance. The slight increase in path

length, inﬂuenced by methodological changes, under-

scores the need for consistent measurement practices

in future recordings and analysis.

Despite the fully automated process from record-

ing the speed climbing run to displaying the data, it

takes a qualiﬁed person who either brings extensive

experience in this sport and/or has a sport scientiﬁc

background to be able to efﬁciently draw conclusions

about errors in movement and performance evalua-

tion. The involvement of a sports scientist, particu-

larly one specialized in applied biomechanics, would

be crucial. Such an expert would collaborate with the

coach responsible for the athlete’s technical prepara-

tion, as well as with the strength and conditioning

trainer. For both coaches, the data would provide

valuable insights for reﬁning and optimizing train-

ing programs, which would ultimately lead to an im-

provement of performance in all sections of the wall.

For future projects, we want to maintain the col-

laboration with the NYSI in order to follow the devel-

opment of young athletes and to gain further insights

into the performance of speed climbing athletes by ex-

panding the data sets and improving the measurement

system.

REFERENCES

Beltr

an, R. B., Richter, J., and Heinkel, U. (2022). Au-

tomated human movement segmentation by means of

human pose estimation in rgb-d videos for climbing

motion analysis.

Cao, Z., Hidalgo, G., Simon, T., Wei, S.-E., and Sheikh,

Y. (2018). Openpose: Realtime multi-person 2d pose

estimation using part afﬁnity ﬁelds.

Chen, K., Wang, J., Pang, J., Cao, Y., Xiong, Y., Li, X.,

Sun, S., Feng, W., Liu, Z., Xu, J., Zhang, Z., Cheng,

D., Zhu, C., Cheng, T., Zhao, Q., Li, B., Lu, X., Zhu,

R., Wu, Y., Dai, J., Wang, J., Shi, J., Ouyang, W., Loy,

C. C., and Lin, D. (2019). MMDetection: Open mm-

lab detection toolbox and benchmark. arXiv preprint

arXiv:1906.07155.

Contributors, M. (2020). Openmmlab pose estimation tool-

box and benchmark.

Elias, P., Skvarlova, V., and Zezula, P. (2021). Speed21:

Speed climbing motion dataset. In Lienhart, R., edi-

tor, Proceedings of the 4th International Workshop on

Multimedia Content Analysis in Sports, ACM Digi-

tal Library, pages 43–50, New York,NY,United States.

Association for Computing Machinery.

Evans, M., Colyer, S., Salo, A., and Cosker, D. (2021). Au-

tomatic high ﬁdelity foot contact location and timing

for elite sprinting. Machine Vision and Applications,

32(5).

Grønhaug, G. and Norberg, M. (2016). First overview

on chronic injuries in sport climbing: proposal for a

change in reporting of injuries in climbing. BMJ Open

Sport & Exercise Medicine, 2:e000083.

Hunter, J. P., Marshall, R. N., and McNair, P. J. (2004).

Segment-interaction analysis of the stance limb in

sprint running. Journal of Biomechanics, 37(9):1439–

1446.

Hunter, J. P., Marshall, R. N., and McNair, P. J. (2005).

Relationships between ground reaction force impulse

and kinematics of sprint-running acceleration. Journal

of Applied Biomechanics, 21(1):31 – 43.

International Federation of Sport Climbing (2024). Speed

world records. Accessed: 2024-10-03.

Legreneur, P., Rogowski, I., and Durif, T. (2019). Kinematic

analysis of the speed climbing event at the 2018 youth

olympic games. Computer Methods in Biomechanics

and Biomedical Engineering, 22(sup1):S264–S266.

Levernier, G., Samozino, P., and Laffaye, G. (2020). Force-

velocity-power proﬁle in high elite boulder, lead,

speed, climbers competitors. International Journal of

Sports Physiology and Performance, 15.

Lin, T., Maire, M., Belongie, S., Bourdev, L., Girshick, R.,

Hays, J., Perona, P., Ramanan, D., Zitnick, C. L., and

Doll

ar, P. (2015). Microsoft coco: Common objects in

context.

Mero, A., Komi, P. V., and Gregor, R. J. (1992).

Biomechanics of sprint running. Sports Medicine,

13(6):376–392.

Pandurevi

c, D., Draga, P., Sutor, A., and Hochradel, K.

(2022). Analysis of competition and training videos

of speed climbing athletes using feature and human

body keypoint detection algorithms. Sensors (Basel,

Switzerland), 22(6).

Reveret, L., Chapelle, S., Quaine, F., and Legreneur, P.

(2020). 3d visualization of body motion in speed

climbing. Frontiers in psychology, 11:2188.

Weyand, P. G., Sternlight, D. B., Bellizzi, M. J., and Wright,

S. (2000). Faster top running speeds are achieved with

greater ground forces not more rapid leg movements.

Journal of applied physiology (Bethesda, Md. : 1985),

89(5):1991–1999.

Winkler, M., K

unzell, S., and Augste, C. (2022). The load

structure in international competitive climbing. Fron-

tiers in Sports and Active Living, 4.

icSPORTS 2024 - 12th International Conference on Sport Sciences Research and Technology Support

240