Sports Analytics Giuseppe Prencipe - Dipartimento di Informatica Università di Pisa - THESES 2019
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Sports Analytics
Giuseppe Prencipe
Dipartimento di Informatica
Università di Pisa
#1
Collect data
#2
Analyse them
5BN/year market
#1 #3
Collect data Automatically
#2
Analyse them
5BN/year market
MAIN SPORT-TECH CLUSTERS Athletic Fan Performance Experience Smart Club Arena Management Immersive E-Sports Media
MAIN SPORT-TECH CLUSTERS Athletic Fan Performance Experience Smart Club Arena Management Immersive E-Sports Media
At our Department PlayeRank
PlayeRank data-driven performance evaluation
ng r r
r d ng i de e r ck de ck
wa t wi ht
w if el eld t ba fiel ba
fo
r
lef id fi lef t ht
rig m lef
t h rig
rig
Collaborations Injury Forecasting
FC Barcelona Done: AI-based Injury Forecaster (IF) for
Spain soccer players
Ferencvárosi TC Ongoing: Explainable AI tool to help
Hungary staff interpret predictions the IF
Philadelphia 76ers Future work: Generator of injury-free
USA
training plans, based on Adversarial
Learning and Generative Adversarial
Univ of Connecticut
USA Networks (GANs)
PlayeRank
Collaborations Tactical Analysis
UEFA Done: AI-based evaluators for teams
Europe and players
Fraunhofer
Bonn, Germany
Ongoing: Explainable AI tools for
Northeastern Univ. forecasting career evolution of players
Boston, USA
and performance of teams
Queen Mary Univ.
London, UK Future work:
FIGC ● Match simulators based on GANs
Italy ● Optimal team formation
● Automatic commentator PlayeRank
11
Video analysis
Live Stroke analysis Live AI data Social network
& Heatmap interpretationProblem 1: position tracking
Video processing, with object
detection + projection on the court,
using keypoints to calibrate
Problem 2: tracking arm’s movements with smartwatch
Classification with supervised learning of the gesture
(data from accelerometer and gyroscope)
Problem 3: coaching
Profile of the player and adaptive expert
system for real-time strategic supportOngoing & Future work: • Effective position tracking for double • Automatic calibration • Pose estimation and ML to track arm’s movements through video • Energy issues • Tackle other sports
E er tennis pla er s pet pee e is ,
Colpi nello smartwatch nderperforming d ring a game, for no
l’app prototipo è italiana apparent reason, has this e er happened to
Dall’Università di Pisa un nuovo
passo avanti nel tennis connesso o ?Stampa Stampa senza immagine Chiudi
J NE 3, 2016 17:47
Pag.22
Colpi nello smartwatch
Anno XII - n.22 - 8 giugno 2016
l’app prototipo è italiana b FEDE ICO CO INI IE 6871
Dall’Università di Pisa un nuovo EA ABLE
passo avanti nel tennis connesso
Tenni , ar no mar a ch ad allenar i per il colpo Pag.22 NE M L IMEDIA ECHNIQ E RANKING LI E
Nole moschettiere
Garbine divina, Nole moschettiere
Eperfe o
er tennis pla er s pet pee e is
Coppa Suzanne Lenglen
alla Muguruza, Coppa Ldei nderperforming
, d, ring a game, for no
Moschettieri a Djokovic:
Coppa Suzanne Lenglen imprese storiche : reason, has this e er happened to
apparent
Pag.4 e 8
o ?
Stampa Stampa senza immagine Chiudi
alla Muguruza, Coppa dei Peppe Aq a o
Moschettieri a Djokovic:
J NE 3, 2016 17:47
EA ABLE
b FEDE ICO CO INI IE 6871
Tenni , ar no mar a ch ad allenar i per il colpo
imprese storiche perfe o
È scientifico: il tennis
L , ,
non fa venire la scoliosi
Pag.4 e 8
Lo dimostra uno studio pubblicato
sulla rivista European Spine Journal : N
Pag.15
Peppe Aq a o
.
Internazionali d’Italia H e Te i Ne O he Ne
Tpra: amatori al Foro
Arrivano in 600 per il SuperSlam
a Roma: le immagini e i risultati
.C
E
' ,
Pag.16
-- ?I . E ! '
Sfida di rovesci: GLI ALTRI CONTENUTI
Prima pagina: Amore e physique du role Pag.3
una mano contro due
Confronto biomeccanico tra i due
Terza pagina: Lilì de Alvarez, prima diva di Spagna
Pag.7 - I numeri della settimana Pag.12 - Il tennis N ENNI COMMANDER:
.M
modi di colpire: vantaggi e svantaggi in TV Pag.14 - Paddle: un circuito... mondiale Pag.19
La regola del gioco: Dopo la sospensione, ,contacts
Giuseppe Prencipe Luca Pappalardo
giuseppe.prencipe@unipi.it luca.pappalardo@unipi.it
Paolo Ferragina Paolo Cintia
paolo.ferragina@unipi.it paolo.cintia@isti.cnr.it
References:
• PlayeRank: data-driven performance evaluation and player ranking in soccer via a machine learning approach.
L. Pappalardo et al. ACM TIST, 2019
• Explainable Injury Forecasting in Soccer via Multivariate Time Series and Convolutional Neural Networks.
L. Pappalardo et al. BARÇA Sports Analytics Summit, 2019
• Who is going to get hurt? Predicting injuries in professional soccer. A. Rossi et al. In Proceedings of MLSA’17You can also read
