Weakly and Semi Supervised Human Body Part Parsing via Pose-Guided Knowledge Transfer

Abstract

Human body part parsing, or human semantic part segmentation, is fundamental to many computer vision tasks. In conventional semantic segmentation methods, the ground truth segmentations are provided, and fully convolutional networks (FCN) are trained in an end-to-end scheme. Although these methods have demonstrated impressive results, their performance highly depends on the quantity and quality of training data. In this paper, we present a novel method to generate synthetic human part segmentation data using easily-obtained human keypoint annotations. Our key idea is to exploit the anatomical similarity among human to transfer the parsing results of a person to another person with similar pose. Using these estimated results as additional training data, our semi-supervised model outperforms its strong-supervised counterpart by 6 mIOU on the PASCAL-Person-Part dataset [6], and we achieve stateof-the-art human parsing results. Our approach is general and can be readily extended to other object/animal parsing task assuming that their anatomical similarity can be annotated by keypoints. The proposed model and accompanying source code will be made publicly available. * This work was done when Xiaolin Fang was an intern at MVIG lab of Shanghai Jiao Tong University.† The corresponding author is Cewu Lu, email: lucewu@sjtu.edu.cn. Cewu Lu is also a member of SJTU-SenseTime lab and AI research institute of SJTU.

A Review of Body Measurement Using 3D Scanning

“…State-of-the-art semantic segmentation methods [33], [49], [93], [96], [171], similar to human pose estimation, are also deep learning-based. In addition to wholebody segmentation [97], [112], there are also body-partsegmented datasets [95], [178].…”

CrowdPose: Efficient Crowded Scenes Pose Estimation and a New Benchmark

“…Estimating multi-person poses in images plays an important role in the area of computer vision. It has attracted tremendous interest for its wide applications in activity understanding [23,15], human re-identification [27], human parsing [31,12] etc. Currently, most of the methods can be roughly divided into two categories: i) top-down approaches that firstly detect each person and then perform single person pose estimation, or ii) bottom-up approaches which detect each joint and then associate them into a whole person.…”

Deep-segmentation of plantar pressure images incorporating fully convolutional neural networks

“…After obtaining the seed region, it is now considered how to use the seed cues to train the image semantic segmentation network. Considering the uneven distribution of foreground and background seed regions, unlike the seed loss proposed in the seeding expansion and constrain loss (SECL) method, the balanced sowing loss proposed by this method (balanced sowing loss), two normalization coefficients with foreground and background respectively encourage the prediction of the segmentation network to match only the seeds [20]. The overall model is shown in Fig.…”