人体关键点SDK
检测图片中所有的行人，并识别每个人的肢体关键点。
SDK功能
- 行人检测
- 肢体关键点检测
SDK包含两个检测器：
- PoseResnet18Estimation
simple pose, backbone: resnet18, dataset:imagenet - PoseResnet50Estimation
simple pose, backbone: resnet50, dataset:imagenet
运行例子 - PoseResnet18EstimationExample
- 测试图片
运行成功后，命令行应该看到下面的信息:
...
[INFO ] -
[ Joint [x=0.563, y=0.141], confidence: 0.9104,
Joint [x=0.646, y=0.109], confidence: 0.8986,
Joint [x=0.521, y=0.109], confidence: 0.9084,
Joint [x=0.792, y=0.125], confidence: 0.8644,
Joint [x=0.479, y=0.109], confidence: 0.8084,
Joint [x=0.875, y=0.359], confidence: 0.5848,
Joint [x=0.396, y=0.234], confidence: 0.5806,
Joint [x=0.896, y=0.609], confidence: 0.4071,
Joint [x=0.021, y=0.266], confidence: 0.7737,
Joint [x=0.979, y=0.781], confidence: 0.4289,
Joint [x=0.313, y=0.094], confidence: 0.7359,
Joint [x=0.708, y=0.750], confidence: 0.4298,
Joint [x=0.375, y=0.703], confidence: 0.4759
]
[INFO ] - [
class: "person", probability: 0.98573, bounds: [x=0.695, y=0.063, width=0.304, height=0.926]
class: "person", probability: 0.90103, bounds: [x=0.000, y=0.021, width=0.310, height=0.930]
class: "person", probability: 0.73586, bounds: [x=0.317, y=0.004, width=0.380, height=0.991]
]
帮助
引擎定制化配置，可以提升首次运行的引擎下载速度，解决外网无法访问或者带宽过低的问题。
引擎定制化配置
官网：
http://www.aias.top/
Github地址：
https://github.com/mymagicpower/AIAS
https://gitee.com/mymagicpower/AIAS