EgoFlow: Gradient-Guided Flow Matching for Egocentric 6DoF Object Motion Generation

TL;DR

EgoFlow is a novel flow-matching framework that generates realistic, physically plausible 6DoF object trajectories from egocentric videos, integrating physical constraints and multimodal observations.

Contribution

It introduces a hybrid architecture combining Mamba, Transformer, and Perceiver models with gradient-guided inference for physically consistent motion synthesis.

Findings

Outperforms diffusion and transformer baselines in accuracy and realism.

Reduces collision rates by up to 79%.

Generalizes well to unseen scenes.

Abstract

Understanding and predicting object motion from egocentric video is fundamental to embodied perception and interaction. However, generating physically consistent 6DoF trajectories remains challenging due to occlusions, fast motion, and the lack of explicit physical reasoning in existing generative models. We present EgoFlow, a flow-matching framework that synthesizes realistic and physically plausible trajectories conditioned on multimodal egocentric observations. EgoFlow employs a hybrid Mamba-Transformer-Perceiver architecture to jointly model temporal dynamics, scene geometry, and semantic intent, while a gradient-guided inference process enforces differentiable physical constraints such as collision avoidance and motion smoothness. This combination yields coherent and controllable motion generation without post-hoc filtering or additional supervision. Experiments on real-world datasets HD-EPIC, EgoExo4D, and HOT3D show that EgoFlow outperforms diffusion-based and transformer baselines in accuracy, generalization, and physical realism, reducing collision rates by up to 79%, and strong generalization to unseen scenes. Our results highlight the promise of flow-based generative modeling for scalable and physically grounded egocentric motion understanding.