Since we haven't decided which OS to choose, I had to have tried both (Win10 and Ubuntu, of course). This document is for environment setting in Linux.
Here is someone else's guide.
RTFM and STFW, and U can do it. Attention, at least 30GB space of hard disk is needed.
PLZ back up your data if you are not familiar with OS installation.
If you are in special regions, PLZ change the downloading source. I recommend using the built-in software Software and Updates(软件与更新）to change the source.
You can find this in the built-in software Software and Updates(软件与更新）.
Normally you need to remove the original gcc, and install v4.8.
sudo apt remove gcc
sudo apt install gcc-4.8PLZ PLZ install CUDA via JetPack 3.3! It's quite safe and sound! Click here to find and download JetPack 3.3!
Anyway you'll need a Nvidia® developer account.
We only need to install CUDA and openCV, so make sure that you have and only have ticked CUDA and openCV.
As for cuDNN, I strongly recommand binary install package.
This step is pretty easy, but also PLZ change download source of pip if you are in special regions.
Official installation guide is here. I strongely recommend choosing the stable one with GPU.
We should use version r1.8.
You need to compile with Bazel 10.
While setting configuration, I recommand to choose 'N' for all except cuda.
https://www.jianshu.com/p/a64bd41c585e
To install Pangolin, you need to install glew first:sudo apt-get install libglew-dev
Pangolin install guide can be seen at the official repo.
Download the archive and unzip it at somewhere.
#How to install a cmake project like nanoflann
cd nanoflann-<version>
mkdir build
cmake ..
make
make installTO BE CONTINUED