Assuming you are looking for a person-attribute prediction, here's an implementation that predicts a wide range of attributes: https://github.com/ZHAOZHIHAO/RunningProgramms/tree/master/running_vespa_pedestrian_attribute. Associated paper: https://arxiv.org/pdf/1707.06089.pdf.
Title: Few-Example Object Detection with Model Communication
Description: With the advent of deep learning based object detection into so many real world applications, it is getting difficult to find annotated datasets for not-so-common objects (eg. street hawkers/vendors, etc). Study of this paper could thus be an interesting gateway for active learning.