3d-pose-2d-keypoints:从2d关键点进行3d人体姿势估计-源码

上传者: 42131352 | 上传时间: 2021-09-18 23:38:34 | 文件大小: 46.61MB | 文件类型: ZIP
从2d关键点进行3d人体姿势估计 概述 尽管人们通常可以轻松地估计2d图像中人的3d姿势,但是3d姿势估计对于机器来说仍然是一个具有挑战性的问题。 该项目改进了一种算法,该算法以2d关键点作为唯一输入来估计人体姿势的3d关键点。 我将采取三种关键干预措施来改善整个数据集以及基准模型中具有特别高误差的姿势子集的重建精度:a)修改预处理中的数据归一化技术,b)从简单密集地修改神经网络架构将网络连接到以最新的2d姿态估计模型为模型的多级网络,并且c)生成合成数据以增强训练集。 这些干预措施成功地将整个测试集(来自卡耐基梅隆大学的运动捕捉数据库)中的重建误差降低了40%,针对目标的高误差姿势也降低了87%。 全文: (pdf) 依存关系 仅用于从头准备数据: 培训与测试 训练 运行prep_data.py 运行train.py ,注释掉所有您不想训练的模型设置。 日期时间将附加到这些文

文件下载

资源详情

[{"title":"( 39 个子文件 46.61MB ) 3d-pose-2d-keypoints:从2d关键点进行3d人体姿势估计-源码","children":[{"title":"3d-pose-2d-keypoints-present","children":[{"title":"preprocessing.py <span style='color:#111;'> 4.17KB </span>","children":null,"spread":false},{"title":".ipynb_checkpoints","children":[{"title":"model_analysis-checkpoint.ipynb <span style='color:#111;'> 3.11MB </span>","children":null,"spread":false}],"spread":true},{"title":"prep_data.py <span style='color:#111;'> 2.01KB </span>","children":null,"spread":false},{"title":"blender","children":[{"title":"save_seated.py <span style='color:#111;'> 1.45KB </span>","children":null,"spread":false},{"title":"save_blender_mirrored.py <span style='color:#111;'> 4.31KB </span>","children":null,"spread":false},{"title":"save_blender_3d_points.py <span style='color:#111;'> 2.35KB </span>","children":null,"spread":false}],"spread":true},{"title":"models.py <span style='color:#111;'> 7.82KB </span>","children":null,"spread":false},{"title":"utils","children":[{"title":"pose_before_after_colors.png <span style='color:#111;'> 108.76KB </span>","children":null,"spread":false},{"title":"pose_before_after.png <span style='color:#111;'> 100.46KB </span>","children":null,"spread":false}],"spread":true},{"title":"README.md <span style='color:#111;'> 3.84KB </span>","children":null,"spread":false},{"title":"Ludlow_3d_pose_2d_keypoints.pdf <span style='color:#111;'> 1.74MB </span>","children":null,"spread":false},{"title":"standardize_bvh.py <span style='color:#111;'> 1.25KB </span>","children":null,"spread":false},{"title":"model_analysis.ipynb <span style='color:#111;'> 3.11MB </span>","children":null,"spread":false},{"title":"logs","children":[{"title":"weights","children":[{"title":"sequence-base-mirror-seated.hdf5 <span style='color:#111;'> 81.85KB </span>","children":null,"spread":false},{"title":"sequence-multi-seated.hdf5 <span style='color:#111;'> 365.94KB </span>","children":null,"spread":false},{"title":"frame-dropout.hdf5 <span style='color:#111;'> 84.88KB </span>","children":null,"spread":false},{"title":"sequence-dropout.hdf5 <span style='color:#111;'> 84.88KB </span>","children":null,"spread":false},{"title":"sequence-multi-mirror.hdf5 <span style='color:#111;'> 366.02KB </span>","children":null,"spread":false},{"title":"sequence-base.hdf5 <span style='color:#111;'> 81.88KB </span>","children":null,"spread":false},{"title":"sequence-multi.hdf5 <span style='color:#111;'> 365.94KB </span>","children":null,"spread":false},{"title":"sequence-multi-seated-short.hdf5 <span style='color:#111;'> 365.94KB </span>","children":null,"spread":false},{"title":"frame-multi.hdf5 <span style='color:#111;'> 366.02KB </span>","children":null,"spread":false},{"title":"frame-base.hdf5 <span style='color:#111;'> 81.85KB </span>","children":null,"spread":false},{"title":"sequence-multi-mirror-seated.hdf5 <span style='color:#111;'> 365.94KB </span>","children":null,"spread":false}],"spread":false},{"title":"train_test","children":[{"title":"Subject14testsequence.npy <span style='color:#111;'> 3.82MB </span>","children":null,"spread":false},{"title":"Subject15testsequence.npy <span style='color:#111;'> 5.00MB </span>","children":null,"spread":false},{"title":"Subject14testframe.npy <span style='color:#111;'> 3.82MB </span>","children":null,"spread":false},{"title":"Subject15testframe.npy <span style='color:#111;'> 5.00MB </span>","children":null,"spread":false},{"title":"Subject13testsequence.npy <span style='color:#111;'> 3.71MB </span>","children":null,"spread":false},{"title":"Subject13testframe.npy <span style='color:#111;'> 3.71MB </span>","children":null,"spread":false}],"spread":true},{"title":"blender_animations","children":[{"title":"seated_animation_short.blend <span style='color:#111;'> 1.50MB </span>","children":null,"spread":false},{"title":"seated_animation.blend <span style='color:#111;'> 1.78MB </span>","children":null,"spread":false}],"spread":true},{"title":"blender_dicts","children":[{"title":"Subject15.p <span style='color:#111;'> 7.41MB </span>","children":null,"spread":false},{"title":"Subject86-mirror.p <span style='color:#111;'> 5.89MB </span>","children":null,"spread":false},{"title":"Subject14.p <span style='color:#111;'> 5.65MB </span>","children":null,"spread":false},{"title":"Subject86.p <span style='color:#111;'> 5.89MB </span>","children":null,"spread":false},{"title":"Subject86-seated.p <span style='color:#111;'> 493.76KB </span>","children":null,"spread":false},{"title":"Subject13.p <span style='color:#111;'> 5.49MB </span>","children":null,"spread":false}],"spread":false}],"spread":true},{"title":"train.py <span style='color:#111;'> 2.79KB </span>","children":null,"spread":false}],"spread":false}],"spread":true}]

评论信息

免责申明

【只为小站】的资源来自网友分享,仅供学习研究,请务必在下载后24小时内给予删除,不得用于其他任何用途,否则后果自负。基于互联网的特殊性,【只为小站】 无法对用户传输的作品、信息、内容的权属或合法性、合规性、真实性、科学性、完整权、有效性等进行实质审查;无论 【只为小站】 经营者是否已进行审查,用户均应自行承担因其传输的作品、信息、内容而可能或已经产生的侵权或权属纠纷等法律责任。
本站所有资源不代表本站的观点或立场,基于网友分享,根据中国法律《信息网络传播权保护条例》第二十二条之规定,若资源存在侵权或相关问题请联系本站客服人员,zhiweidada#qq.com,请把#换成@,本站将给予最大的支持与配合,做到及时反馈和处理。关于更多版权及免责申明参见 版权及免责申明