Learning to Detect and Track Visible and Occluded Body Joints in a Virtual World

TL;DR

This paper introduces a deep learning architecture for multi-person tracking that jointly detects visible and occluded body parts, leveraging a large synthetic dataset to improve accuracy in urban scene tracking.

Contribution

The authors propose a novel end-to-end network with four branches for joint detection and association of body parts, trained on the largest synthetic dataset for urban human tracking.

Findings

Model effectively detects occluded and visible joints.

Architecture generalizes well to real-world benchmarks.

Synthetic data enables robust tracking in urban scenarios.

Abstract

Multi-People Tracking in an open-world setting requires a special effort in precise detection. Moreover, temporal continuity in the detection phase gains more importance when scene cluttering introduces the challenging problems of occluded targets. For the purpose, we propose a deep network architecture that jointly extracts people body parts and associates them across short temporal spans. Our model explicitly deals with occluded body parts, by hallucinating plausible solutions of not visible joints. We propose a new end-to-end architecture composed by four branches (visible heatmaps, occluded heatmaps, part affinity fields and temporal affinity fields) fed by a time linker feature extractor. To overcome the lack of surveillance data with tracking, body part and occlusion annotations we created the vastest Computer Graphics dataset for people tracking in urban scenarios by exploiting a photorealistic videogame. It is up to now the vastest dataset (about 500.000 frames, almost 10 million body poses) of human body parts for people tracking in urban scenarios. Our architecture trained on virtual data exhibits good generalization capabilities also on public real tracking benchmarks, when image resolution and sharpness are high enough, producing reliable tracklets useful for further batch data association or re-id modules.