Sequential Attend, Infer, Repeat: Generative Modelling of Moving Objects

TL;DR

This paper introduces SQAIR, a deep generative model for videos that can detect, track, and generate moving objects over time, improving upon previous models by handling occlusions and applying to real-world data.

Contribution

SQAIR extends AIR by incorporating temporal dynamics, enabling unsupervised detection, tracking, and future frame generation of moving objects in videos.

Findings

SQAIR outperforms AIR in detecting overlapping and occluded objects.

SQAIR successfully tracks pedestrians in CCTV footage without supervision.

The model can generate realistic future frames based on current observations.

Abstract

We present Sequential Attend, Infer, Repeat (SQAIR), an interpretable deep generative model for videos of moving objects. It can reliably discover and track objects throughout the sequence of frames, and can also generate future frames conditioning on the current frame, thereby simulating expected motion of objects. This is achieved by explicitly encoding object presence, locations and appearances in the latent variables of the model. SQAIR retains all strengths of its predecessor, Attend, Infer, Repeat (AIR, Eslami et. al., 2016), including learning in an unsupervised manner, and addresses its shortcomings. We use a moving multi-MNIST dataset to show limitations of AIR in detecting overlapping or partially occluded objects, and show how SQAIR overcomes them by leveraging temporal consistency of objects. Finally, we also apply SQAIR to real-world pedestrian CCTV data, where it learns to reliably detect, track and generate walking pedestrians with no supervision.