Person Recognition in Personal Photo Collections

TL;DR

This paper introduces a simple yet effective person recognition framework leveraging convnet features from multiple regions, achieving state-of-the-art results on social media photo datasets despite challenges like viewpoint and appearance variations.

Contribution

The paper presents a new person recognition approach that combines multiple image regions and analyzes feature importance for generalizability, setting new benchmarks on social media photo datasets.

Findings

Achieves state-of-the-art results on the PIPA benchmark.

Demonstrates the effectiveness of multi-region convnet features.

Provides detailed analysis of feature importance and viewpoint robustness.

Abstract

People nowadays share large parts of their personal lives through social media. Being able to automatically recognise people in personal photos may greatly enhance user convenience by easing photo album organisation. For human identification task, however, traditional focus of computer vision has been face recognition and pedestrian re-identification. Person recognition in social media photos sets new challenges for computer vision, including non-cooperative subjects (e.g. backward viewpoints, unusual poses) and great changes in appearance. To tackle this problem, we build a simple person recognition framework that leverages convnet features from multiple image regions (head, body, etc.). We propose new recognition scenarios that focus on the time and appearance gap between training and testing samples. We present an in-depth analysis of the importance of different features according to time and viewpoint generalisability. In the process, we verify that our simple approach achieves the state of the art result on the PIPA benchmark, arguably the largest social media based benchmark for person recognition to date with diverse poses, viewpoints, social groups, and events. Compared the conference version of the paper, this paper additionally presents (1) analysis of a face recogniser (DeepID2+), (2) new method naeil2 that combines the conference version method naeil and DeepID2+ to achieve state of the art results even compared to post-conference works, (3) discussion of related work since the conference version, (4) additional analysis including the head viewpoint-wise breakdown of performance, and (5) results on the open-world setup.