Confused about coding on Jetson

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hello, I just recently received a Jetson TX2 for my birthday and I’m trying to use the onboard camera for simple image detection.

I’m trying to follow these 2 tutorials here:
[url]https://github.com/dusty-nv/jetson-inference#system-setup[/url]
[url]https://github.com/KleinYuan/tx2-flash[/url]

where it involves using a host PC and installing JetPack to flash.

I am wondering if there is a simple tutorial using Caffe/TensorFlow on the Jetson itself without the need for a host PC? I’m a bit new to all this and would appreciate your help.

Thank you.

2

I couldn’t tell you about the Caffe/TensorFlow tutorial, but flashing the Jetson requires any x86_64 Linux PC host…using JetPack (which is a front end) implies it also needs to be Ubuntu 14.04 or 16.04 (JetPack understands Ubuntu packaging). So if you have a system with something like Windows on it, then you might consider making it dual boot…the only downside to this is that you’ll need plenty of disk space…it’s easier to add a second disk than it is to split an existing disk (flashing takes a lot of space).

3

Moreover, once you have flashed your Jetson you may try several things including tensorflow installation with inception and out of-the box image detection solution GitHub - dusty-nv/jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.

4

The Jetson may come pre-flashed with some version of JetPack.
If so, you may not HAVE to flash from a host PC.
If you have to flash from a host PC, using a USB stick to boot a PC into Ubuntu 16.04 is typically the easiest way to get that going.

If your module already has a JetPack on it, you can probably skip all the way to https://github.com/dusty-nv/jetson-inference#building-from-source-on-jetson

I also think people have been able to build and run DarkNet, with the Yolo or Yolo-tiny model.
Those are pretty good pre-trained models for object detection.

5

For the latter refer to https://developer.nvidia.com/embedded/twodaystoademo & yolo.

6 
It seems he is not a engineer,may be it's too dificullt for he to use ubuntu,and some usb device's storage space are not enough for jetson tx2's flashing,so he may be need to use "mount" command .....
I have no idea about:a new tx2 board,do It been installed Caffe/TensorFlow package.
By the way,it is a good idea to use the lastest packages,so the board need a flashing.
find a friend who use ubuntu16.04,and use his computer to flashing the board,the network station should be good!
plug a keyboard,a mouse,a screen,tx2 just a computer!