Throwing4: Phase-Aware Pose Modeling for Throw-Distance Prediction in Elite Athletics
Authors: Bikash Kumar Badatya, Kartike Tiwari, Jyotirmoy Amin, Ravi Hegde
Organizations: Department of Electrical Engineering, Indian Institute of Technology Gandhinagar, Gandhinagar, India. · Department of Electronics and Communication Engineering, Madan Mohan Malaviya University of Technology, Gorakhpur, India. · Department of Mathematical Sciences, IISER Berhampur, Berhampur, India.
Elite athletics throwing provides a structured biomechanical action sequence together with an objective continuous outcome, namely throw distance, making it a suitable setting for performance-oriented sports analytics. In this work, we investigate whether explicit phase alignment of pose dynamics improves throw-distance prediction under realistic broadcast conditions. To support this study, we curate Throwing4, a dataset of 500 elite throwing clips spanning javelin, discus, hammer, and shot put, with corresponding two-dimensional pose sequences, frame-level phase annotations, and official distance labels. We formulate throw-distance prediction from broadcast-derived pose as a regression problem and compare full-sequence modeling against a phase-conditioned approach that partitions each attempt into four technique stages and fuses phase-specific motion representations. Experiments across multiple graph-based pose encoders show a consistent reduction in prediction error when phase structure is incorporated, indicating that phase-aligned modeling provides a more stable representation than treating the entire motion as an unstructured sequence. These findings suggest that explicit biomechanical phase information is a useful inductive prior for pose-based performance modeling in elite athletics throwing. Beyond score prediction, the dataset and protocol provide a basis for future work on fine-grained technique analysis, interpretable sports analytics, and robust modeling from broadcast video.
Figures & tables
Sport
Videos (M/F)
Tokyo
Paris
WAC
EAC
Phase
2020
2024
Beijing 2015
Oregon 2022
2024
1
2
3
4
Javelin Throw
211 (111/100)
45
82
27
16
41
step
drive
throw
recovery
Discus Throw
112 (69/43)
60
52
–
–
–
swing
turn
throw
recovery
Hammer Throw
90 (30/60)
26
64
–
–
–
swing
turn
throw
recovery
Shot Put
87 (55/32)
32
43
–
–
12
stance
glide
throw
recovery
Table 1: Throwing4 overview: number of clips per sport with male/female split (Videos (M/F)), source broadcast events, and the canonical four-phase taxonomy used for temporal annotation.
Figure 1: Qualitative illustration of the four-phase technique progression for representative clips from javelin, discus, hammer, and shot put.
Figure 2: Pose extraction pipeline in Throwing4. Broadcast RGB clips are processed with MMPose for multi-person detection, followed by identity-consistent thrower tracking to obtain an MPII-16 2D pose sequence for each clip.
Figure 3: Phase-conditioned throw-distance prediction pipeline. The pose sequence is divided into four phases, each phase is encoded by a spatio-temporal graph network, pooled with masking, and fused for regression of the normalized throw distance.
Model
Javelin
Discus
Hammer
Shot put
full
phase
full
phase
full
phase
full
phase
GCN Kipf and Welling (2016)
0.2863
0.2604
0.1278
0.1160
0.2334
0.1680
0.3076
0.2055
STGCN Yan et al. (2018)
0.2583
0.1902
0.1494
0.1317
0.1609
0.1432
0.3651
0.3196
ASTGCN Guo et al. (2019)
0.2023
0.1070
0.1352
0.1249
0.2645
0.1364
0.3976
0.2972
AGCN Shi et al. (2019)
0.0682
0.0662
0.1631
0.1467
0.1686
0.1504
0.3779
0.2191
TSAGCN Shi et al. (2020)
0.1203
0.1034
0.1461
0.1324
0.1559
0.1461
0.4112
0.2213
Table 2: MAE in normalized space for full-sequence and phase-conditioned inputs across four throwing disciplines. Lower is better.
Figure 4: MAE in meters for full-sequence and phase-conditioned score prediction across four throwing disciplines. Errors in meters are obtained by de-normalizing the normalized MAE using Eq. ( 11 ).
How much can a pitcher's body reveal about the upcoming pitch? We study this question at scale by classifying eight pitch types from monocular 3D pose sequences, without access to ball-flight data. Our pipeline chains a diffusion-based 3D pose backbone with automatic pitching-event detection, groundtruth-validated biomechanical feature extraction, and gradient-boosted classification over 229 kinematic features. Evaluated on 119,561 professional pitches, the largest such benchmark to date, we achieve 80.4% accuracy using body kinematics alone. A systematic importance analysis reveals that upper-body mechanics contribute 64.9% of the predictive signal versus 35.1% for the lower body, with wrist position (14.8%) and trunk lateral tilt emerging as the most informative joint group and biomechanical feature, respectively. We further show that grip-defined variants (four-seam vs.\ two-seam fastball) are not separable from pose, establishing an empirical ceiling near 80% and delineating where kinematic information ends and ball-flight information begins.
Jerrin Bright, Michelle Lu, John Zelek
Vision and Image Processing Lab · University of Waterloo, Canada
Injury prediction in pitching depends on precise biomechanical signals, yet gold-standard measurements come from expensive, stadium-installed multi-camera systems that are unavailable outside professional venues. We present a monocular video pipeline that recovers 18 clinically relevant biomechanics metrics from broadcast footage, positioning pose-derived kinematics as a scalable source for injury-risk modeling. Built on DreamPose3D, our approach introduces a drift-controlled global lifting module that recovers pelvis trajectory via velocity-based parameterization and sliding-window inference, lifting pelvis-rooted poses into global space. To address motion blur, compression artifacts, and extreme pitching poses, we incorporate a kinematics refinement pipeline with bone-length constraints, joint-limited inverse kinematics, smoothing, and symmetry constraints to ensure temporally stable and physically plausible kinematics. On 13 professional pitchers (156 paired pitches), 16/18 metrics achieve sub-degree agreement (MAE <1∘). Using these metrics for injury prediction, an automated screening model achieves AUC 0.811 for Tommy John surgery and 0.825 for significant arm injuries on 7,348 pitchers. The resulting pose-derived metrics support scalable injury-risk screening, establishing monocular broadcast video as a viable alternative to stadium-scale motion capture for biomechanics.
Jerrin Bright, Justin Mende, John Zelek
Vision and Image Processing Lab · University of Waterloo, Canada
Trampoline gymnastics involves extreme human poses and uncommon viewpoints, on which state-of-the art pose estimation models tend to under-perform. We demonstrate that this problem can be addressed by fine-tuning a pose estimation model on a combination of real extreme poses and domain-specific synthetic poses (STP). We generate STP from motion capture recordings of trampoline routines. We propose a pipeline to fit noisy motion capture data to a parametric human model, then generate multi-view realistic images with high-fidelity keypoint labels. The fine-tuned ViTPose model tested on real multi-view images exhibits accuracy improvements in 2D which translate to improved 3D triangulation. In 2D, we obtain a performance similar to state-of-the-art models on the MS COCO validation set while evaluating on significantly more challenging data, bridging the performance gap between common and extreme poses. In 3D, we reduce the MPJPE by 46.1 mm with our best model, which represents an improvement of 42.7% compared to the pretrained ViTPose model. Our code and data are available at https://github.com/VisionICLab/trampoline_syn_data.
Léa Drolet-Roy, Victor Nogues, Bérenger Chedal-Anglay +4
Polytechnique Montreal, Canada · Institut National du Sport du Québec, Canada · University of Sherbrooke, Canada +1