RealSense 3D Perception Powers SHOTLAB’s Human Movement Intelligence for Basketball Training
The challenge
SHOTLAB wanted its AI Basketball Training System to explain how a shot was performed and what an athlete could change to make the next one better. Coaches and athletes could already record video and track makes and misses, but conventional 2D video could not fully capture the depth and spatial relationships needed to quantify how joints move in 3D space and how that movement connects to the ball’s trajectory.
The solution
SHOTLAB integrated RealSense D455f depth cameras as the 3D perception layer of its AI Basketball Training System, capturing both athlete movement and ball trajectory. RealSense supplies the continuous depth data SHOTLAB needs to analyze the athlete and the ball within the same spatial and temporal framework, so its engineers can focus on body modeling, comparison against target movements and real-time feedback.
The results
RealSense-powered perception lets SHOTLAB turn each shot into measurable data on mechanics, ball trajectory and outcome, then deliver real-time feedback athletes and coaches can act on. The portable system is already operating at meaningful scale across a wide range of real-world basketball settings, and it can measure whether an athlete’s movement improves through repeated practice.
Stable Tracking Through Occlusion, Motion and Changing Light
Live sports environments bring multiple people in frame, a fast-moving ball, occlusion and changing light. SHOTLAB’s portable hardware defines a clear capture area around the active athlete, while dedicated tracking algorithms built on RealSense’s high-frame-rate, wide-field-of-view depth data keep athlete and ball tracking stable as conditions shift.
Athlete and Ball, Measured Together in 3D
In a typical setup, one D455f captures the athlete’s full-body shooting motion while a second, mounted near the basket, tracks the ball, and SHOTLAB processes and combines both streams locally. SHOTLAB turns that depth data into joint positions, movement phases and shooting-mechanics metrics, then compares them against a target movement to deliver real-time feedback.
A Portable Platform Built on a Mature Perception Layer
Building on RealSense let SHOTLAB put its engineering effort into a reusable movement-learning stack instead of building a sensing stack from scratch, shortening the path from development to deployable products. The resulting system is portable and works wherever athletes practice, without the fixed installation, complex calibration and dedicated motion-capture environment that traditional multi-camera systems require.
“Traditional movement assessment relies heavily on visual observation and coaching expertise, making it difficult to quantify how a movement is executed, how it differs from a target movement, and how consistently it is repeated. We make those aspects measurable, helping athletes and coaches identify specific adjustments, track improvement, and provide quantitative feedback that supports the development and refinement of muscle memory.”
— Xi Yang,
CEO, SHOTLAB
