43998213

> 是的，我在 20.04 jetson 上运行了相同的脚本， > 让我们看看它是否适用于这些脚本：[https](https://github.com/cielavenir/openrave-installation/archive/refs/heads/patch-1.zip) : [//github.com/cielavenir/openrave-installation/archive/refs/heads/patch-1.zip](https://github.com/cielavenir/openrave-installation/archive/refs/heads/patch-1.zip) good job, wish you success

> it seems on 20.04 the script installation got through > osg did not getr through though > but seems all built finally 100% I tried upgrade to 20.04 ,bu...

> you would have to use do-release-upgrade > any jetson including nano will do, but the latter will be the slowest Our country's network may be constrained ，they told me...

> Hello, > > You need to use my fork of pcl when compiling drl: > > * https://github.com/carlosmccosta/pcl > > Check documentation for more information: > > * https://github.com/carlosmccosta/dynamic_robot_localization#installation...

> Hello, > > In the log it shows includes from the /usr/include/pcl-1.8 installed from apt (this mixing of includes from /usr and my fork is causing this problem). >...

> 你要识别的话可以只传一张照片 > 训练的话二者都要成千上万的照片 第一个问题似乎我说的不清晰，我用过百度和虹软人脸识别的SDK，只需要传一张人脸上去，似乎不需要训练，马上就可以识别，事实上所有的人脸识别都是这样的，我想知道目标检测能做到吗？比如有10个手机的类型，我又传了1张手机照片，也能像人脸识别一样，不需要训练就能够识别吗？ 第二个问题用几十个样本训练像这里面说的可行吗？https://www.zhihu.com/question/308399039/answer/573521899 他似乎只用了50个样本，这样准确度会不会有影响，当然了可以直接试一下，但如果你尝试过或许可以直接给我个答案，毕竟我的电脑训练一次要好久呵呵 另外...B站有评论说你‘可爱’，嗯.. 有同感，不过我觉得你更年少有为，加油..! 非常感谢你的答疑

> 1、人脸识别用的算法不一样，facenet都是训练好的，你用的是训练好的模型去预测，而非你用一张图片去训练 优秀.. 又涨姿势了，谢谢