人工智能培训

搜索

人工智能论文:DENSEBODY:从单色图像直接回归密集的3D人体姿势和形状(DenseBody: Directly Regressing Dense 3D H

[复制链接]
cbz123 发表于 2019-3-26 11:45:27 | 显示全部楼层 |阅读模式
cbz123 2019-3-26 11:45:27 393 0 显示全部楼层
人工智能论文:DENSEBODY:从单色图像直接回归密集的3D人体姿势和形状(DenseBody: Directly Regressing Dense 3D Human Pose and Shape From a  Single Color Image)从2D图像恢复3D人体形状和姿势是一项具有挑战性的任务,因为人体的高复杂性和灵活性以及相对较少的3D标记数据。解决这些问题的先前方法通常依赖于预测中间结果,例如身体部位分割,2D / 3D关节,轮廓掩模,以将问题分解为多个子任务,以便利用更多2D标签。大多数先前的工作在他们的方法中结合参数体形状模型并且预测低维空间中的参数以表示人体。在本文中,我们建议使用卷积神经网络(CNN)从单个彩色图像直接回归3Dhuman网格。我们使用3D人体形状和姿势的有效表示,可以通过编码器 - 解码器神经网络预测。所提出的方法在几个3D人体数据集上实现了最先进的性能,包括Human3.6M,SURREAL和UP-3D,具有更快的运行速度。
Recovering 3D human body shape and pose from 2D images is a challenging taskdue to high complexity and flexibility of human body, and relatively less 3Dlabeled data.Previous methods addressing these issues typically rely onpredicting intermediate results such as body part segmentation, 2D/3D joints,silhouette mask to decompose the problem into multiple sub-tasks in order toutilize more 2D labels.Most previous works incorporated parametric body shapemodel in their methods and predict parameters in low-dimensional space torepresent human body.In this paper, we propose to directly regress the 3Dhuman mesh from a single color image using Convolutional Neural Network(CNN).We use an efficient representation of 3D human shape and pose which can bepredicted through an encoder-decoder neural network.The proposed methodachieves state-of-the-art performance on several 3D human body datasetsincluding Human3.6M, SURREAL and UP-3D with even faster running speed.人工智能论文:DENSEBODY:从单色图像直接回归密集的3D人体姿势和形状(DenseBody: Directly Regressing Dense 3D Human Pose and Shape From a  Single Color Image) C97elb0ol8d0P4Q8.jpg
URL地址:https://arxiv.org/abs/1903.10153     ----pdf下载地址:https://arxiv.org/pdf/1903.10153    ----人工智能论文:DENSEBODY:从单色图像直接回归密集的3D人体姿势和形状(DenseBody: Directly Regressing Dense 3D Human Pose and Shape From a  Single Color Image)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

cbz123当前离线
新手上路

查看:393 | 回复:0

快速回复 返回顶部 返回列表