Diversifying Human Pose in Synthetic Data for Aerial-view Human Detection

TL;DR

This paper introduces SynPoseDiv, a framework that enhances synthetic aerial-view human datasets by diversifying poses using diffusion models and image translation, leading to improved detection accuracy especially in low-data scenarios.

Contribution

The paper presents a novel pose diversification method for synthetic data using diffusion-based generation and image translation, addressing pose variety challenges in aerial human detection datasets.

Findings

Significantly improves detection accuracy across benchmarks.

Effective in low-shot training scenarios.

Robust regardless of dataset size or training approach.

Abstract

Synthetic data generation has emerged as a promising solution to the data scarcity issue in aerial-view human detection. However, creating datasets that accurately reflect varying real-world human appearances, particularly diverse poses, remains challenging and labor-intensive. To address this, we propose SynPoseDiv, a novel framework that diversifies human poses within existing synthetic datasets. SynPoseDiv tackles two key challenges: generating realistic, diverse 3D human poses using a diffusion-based pose generator, and producing images of virtual characters in novel poses through a source-to-target image translator. The framework incrementally transitions characters into new poses using optimized pose sequences identified via Dijkstra's algorithm. Experiments demonstrate that SynPoseDiv significantly improves detection accuracy across multiple aerial-view human detection benchmarks, especially in low-shot scenarios, and remains effective regardless of the training approach or dataset size.