Detection of human body parts on the image using the neural networks and the attention model

Abstract

Human body parts detection is a challenging task, which has a lot of applications. In this paper, we propose an algorithm to detect human body parts on images using the OpenPose neural network and the attention model. The novelty of the proposed algorithm is that it is based on a convolutional neural network that uses non-parametric representation to associate the body parts with people in an image in combination with the attention model that learns to focus on specific regions of the input image. The algorithm is part of the Smart Cropping system developed by the authors with the aim to cut necessary pieces of clothing in images and prepare e-commerce catalogues.