TrajLoom: Dense Future Trajectory Generation from Video

TL;DR

TrajLoom introduces a novel framework for dense future trajectory prediction from videos, significantly extending prediction horizons and enhancing motion realism, thereby advancing controllable video generation and editing.

Contribution

It presents a new method combining grid-anchor encoding, a variational autoencoder, and flow matching, along with a unified benchmark for future trajectory prediction.

Findings

Extends prediction horizon from 24 to 81 frames.

Improves motion realism and stability across datasets.

Supports downstream video generation and editing.

Abstract

Predicting future motion is crucial in video understanding and controllable video generation. Dense point trajectories are a compact, expressive motion representation, but modeling their future evolution from observed video remains challenging. We propose a framework that predicts future trajectories and visibility from past trajectories and video context. Our method has three components: (1) Grid-Anchor Offset Encoding, which reduces location-dependent bias by representing each point as an offset from its pixel-center anchor; (2) TrajLoom-VAE, which learns a compact spatiotemporal latent space for dense trajectories with masked reconstruction and a spatiotemporal consistency regularizer; and (3) TrajLoom-Flow, which generates future trajectories in latent space via flow matching, with boundary cues and on-policy K-step fine-tuning for stable sampling. We also introduce TrajLoomBench, a unified benchmark spanning real and synthetic videos with a standardized setup aligned with video-generation benchmarks. Compared with state-of-the-art methods, our approach extends the prediction horizon from 24 to 81 frames while improving motion realism and stability across datasets. The predicted trajectories directly support downstream video generation and editing. Code, model checkpoints, and datasets are available at https://trajloom.github.io/.