Validity and Reliability of an AI-Based Jump Height App vs. Infrared Contact Mat: Minimal Influence of Skin Pigmentation Under Standardized Lighting

Pablo Tadeo Ríos-Gallardo; Luis Enrique Carranza-García; Martin Dietze-Hermosa; Matthew P Gonzalez; Carlos Balsalobre-Fernández; Sandor Dorgo; Samuel Montalvo

doi:10.1519/JSC.0000000000005433

Validity and Reliability of an AI-Based Jump Height App vs. Infrared Contact Mat: Minimal Influence of Skin Pigmentation Under Standardized Lighting

J Strength Cond Res. 2026 Mar 3. doi: 10.1519/JSC.0000000000005433. Online ahead of print.

Authors

Pablo Tadeo Ríos-Gallardo¹, Luis Enrique Carranza-García¹, Martin Dietze-Hermosa², Matthew P Gonzalez³, Carlos Balsalobre-Fernández⁴, Sandor Dorgo³, Samuel Montalvo⁵

Affiliations

¹ School of Sport Organization, Autonomous University of Nuevo Leon, Mexico.
² Department of Human Performance and Recreation, Brigham Young University Idaho, Rexburg, Idaho.
³ Department of Kinesiology, University of Texas at San Antonio, San Antonio, Texas.
⁴ Applied Biomechanics and Sports Technology Research Group, Autonomous University of Madrid, Madrid, Spain; and.
⁵ Wu Tsai Human Performance Alliance, Division of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, California.

PMID: 41875017
DOI: 10.1519/JSC.0000000000005433

Abstract

Ríos-Gallardo, PT, Carranza-García, LE, Dietze-Hermosa, M, Gonzalez, MP, Balsalobre-Fernández, C, Dorgo, S, and Montalvo, S. Validity and reliability of an AI-based jump height app vs. infrared contact mat: minimal influence of skin pigmentation under standardized lighting. J Strength Cond Res XX(X): 000-000, 2026-This study examined the validity, reliability, and visual robustness of an artificial intelligence-based mobile application (My Jump Lab) for measuring countermovement jump (CMJ) and squat jump (SJ) height across a heterogeneous athletic population. A dual-session, test-retest design was implemented with 43 recreationally active adults (age: 21.2 ± 2.4 years), who performed 3 maximal-effort SJ and CMJ trials per session. Jump height was concurrently recorded using a force platform and the AI-based app. Validity was assessed through linear regression and Bland-Altman analyses, and reliability was evaluated using intraclass correlation coefficients (ICC) and coefficient of variation (CV). A linear mixed-effects model tested whether body dimensions, lighting conditions, or skin pigmentation influenced AI accuracy. The AI systematically overestimated jump height (bias = +2.81 cm, p < 0.001), yet showed excellent concurrent validity (R2 = 0.94), strong within-session reliability (ICC = 0.97; CV = 4.2%), and good between-session reliability (ICC = 0.89). Countermovement jump values were more consistent than SJ. No significant effects were observed for lighting or pigmentation (p > 0.05). Although absolute error was higher in SJ, AI-based estimates remained stable across conditions. The level of significance was set at p ≤ 0.05. In conclusion, the AI-based app provides a valid and reliable alternative for field-based jump assessment. However, practitioners should interpret absolute values cautiously, especially for SJ. These findings support the utility of computer vision and AI to democratize biomechanical assessments without sacrificing measurement quality.

Keywords: computer vision; explosive strength; individual typology angle; markless analysis; neuromuscular performance; vertical jump.