slam:RTABSlam方法,实时的slam 3d重建开源方法

本文介绍 slam:RTABSlam方法,实时的slam 3d重建开源方法 slam:RTABSlam方法,实时的slam 3d重建开源方法 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 由于最近在玩RGBD摄像头,对图像分割的结果进行点云映射,得到3D实际场景下的3D点云。……

阅读全文

MobileNetV2,DenseNet,ShuffleNet几种轻量网络性能对比

本文介绍 MobileNetV2,DenseNet,ShuffleNet几种轻量网络性能对比 MobileNetV2,DenseNet,ShuffleNet几种轻量网络性能对比 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 这篇文章简要的对比一下常用的轻量网络之间的性能差……

阅读全文

超分辨率SuperResolution手记一

本文介绍 超分辨率SuperResolution手记一 超分辨率SuperResolution手记一 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 挖个坑,有时间研究一下,这玩意是一个很好的应用场景。并且,许多场合可能会用到。比如重构超高的分辨率图像,甚至是在数据集构建场……

阅读全文

从shuffleseg以及分组卷积说起

本文介绍 从shuffleseg以及分组卷积说起 从shuffleseg以及分组(群)卷积说起 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 什么是shufflenet shufflenet两个核心的骚操作其实就是:逐点的群卷积与通道混洗,in english it is: pointwise group convolution and channel shuffle, so this is how the name……

阅读全文

ROS武功秘籍

本文介绍 ROS武功秘籍 ROS武功秘籍 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 这是ros的一系列记录,其中包括,从0到自己做一个仿真机器人的所有流程。包括自己录制地图,使用地图,进行导航, 机器人控制,局部规划,全局规划,避障算法等一系列应用。 导航 导航部分包含了整……

阅读全文

ndt配准,建图,与无人车定位

本文介绍 ndt配准,建图,与无人车定位 ndt配准,建图,与无人车定位 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 无人车系统中的定位问题 不管是机器人还是无人车,机器人所在的位置,程序是无法知道的。为什么?很简单,因为我无法用GPS或者其他传感器提供定位信息,一句话,……

阅读全文

从为什么VOC的分割标注8位单通道确实彩色图片说起

本文介绍 从为什么VOC的分割标注8位单通道确实彩色图片说起 从为什么VOC的分割标注8位单通道确实彩色图片说起 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 记录一个很基础的问题,却从来没有人关心过。甚至很多人不懂。但是这个问题却不是一个简单的问题,如果你没有弄懂这个问……

阅读全文

点云与图像融合的3D物体检测

本文介绍 点云与图像融合的3D物体检测 点云与图像融合的3D物体检测 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu 做3D检测可以采用纯图像,纯点云,也可以融合点云和图像,毫无疑问,后者是最复杂的。现有的一些比如MultiView或者AVOD都是将点云提取出更多的调整图作……

阅读全文

深度可分离卷积的具体理解:MobileNetV2设计灵感

本文介绍 深度可分离卷积的具体理解:MobileNetV2设计灵感 深度可分离卷积的具体理解:MobileNetV2设计灵感 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu Depthwise separable convolution 在对比不同网络的结构的时候,很多网络它背后的设计思想却很容易被人忽视。其中MobileNet中……

阅读全文

目标检测的前世今身

本文介绍 目标检测的前世今身 目标检测的前世今身 This article was original written by Jin Tian, welcome re-post, first come with https://jinfagang.github.io . but please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu Faster-RCNN算法再思考 实际上2阶段检测算法从现在看来,已经有点过时,但是不失为一种有效的检测方法。对比单阶段检测算法来说,毫无疑问,其准确率和精度都是可以得到保证的。……

阅读全文