The objective of human pose estimation is to estimate the locations of keypoints on the human body using a single image. Convolutional pose machines is one of the most popular pose estimation techniques that is based on deep learning with convolutional features. In this paper, we propose semantic pose machines, a pose estimation technique that enhances convolutional pose machines by utilizing a semantic segmentation heatmap in addition to convolutional features. Semantic segmentation methods leverage the success of object class recog-nition networks for the segmentation of important object classes, including people. We consider the CRF as RNN semantic seg-mentation approach to obtain a heatmap that is incorporated in the pose estimation process as an additional channel. Our results on the LEEDS dataset indicate improvements over the convolutional pose machines method.
Semantic segmentation, classifying each pixel in an image to a set of various objects, is an important and necessary problem to understand images. In recent years, convolutional neural networks trained with public datasets enable to segment objects and understand images. However, it is still challenging to segment objects with high accuracy on a simple and small network. In this work, we describe convolutional neural networks with dilated convolutions to segment person accurately especially near boundary using data augmentation technique. Additionally, we develop a smaller network which can run each frame in webcam video faster without degrading segmentation performance. Our method both numerically and visually outperforms other segmentation techniques.