MAMMA: Markerless & Automatic Multi-Person Motion Action Capture
By: Hanz Cuevas-Velasquez , Anastasios Yiannakidis , Soyong Shin and more
Potential Business Impact:
Captures people's movements without cameras on them.
We present MAMMA, a markerless motion-capture pipeline that accurately recovers SMPL-X parameters from multi-view video of two-person interaction sequences. Traditional motion-capture systems rely on physical markers. Although they offer high accuracy, their requirements of specialized hardware, manual marker placement, and extensive post-processing make them costly and time-consuming. Recent learning-based methods attempt to overcome these limitations, but most are designed for single-person capture, rely on sparse keypoints, or struggle with occlusions and physical interactions. In this work, we introduce a method that predicts dense 2D surface landmarks conditioned on segmentation masks, enabling person-specific correspondence estimation even under heavy occlusion. We employ a novel architecture that exploits learnable queries for each landmark. We demonstrate that our approach can handle complex person--person interaction and offers greater accuracy than existing methods. To train our network, we construct a large, synthetic multi-view dataset combining human motions from diverse sources, including extreme poses, hand motions, and close interactions. Our dataset yields high-variability synthetic sequences with rich body contact and occlusion, and includes SMPL-X ground-truth annotations with dense 2D landmarks. The result is a system capable of capturing human motion without the need for markers. Our approach offers competitive reconstruction quality compared to commercial marker-based motion-capture solutions, without the extensive manual cleanup. Finally, we address the absence of common benchmarks for dense-landmark prediction and markerless motion capture by introducing two evaluation settings built from real multi-view sequences. We will release our dataset, benchmark, method, training code, and pre-trained model weights for research purposes.
Similar Papers
Markerless Tracking-Based Registration for Medical Image Motion Correction
Image and Video Processing
Cleans up X-ray videos to better see how people swallow.
Ego-Exo 3D Hand Tracking in the Wild with a Mobile Multi-Camera Rig
CV and Pattern Recognition
Tracks hands in 3D, even when moving freely.
EchoMask: Speech-Queried Attention-based Mask Modeling for Holistic Co-Speech Motion Generation
Graphics
Makes computer-generated people move like real humans.