In this lesson I will show you how to install few popular pre-trained neural network for Object Detection and Image recognition. There are bunch of neural network which you can start using like ssd-mobilenet, ssd inception, facenet, pednet, googlenet and so on.
All these pre-trained neural network are compiled into Jetson Inference package by NVIDIA for new comers in Artificial Intelligence field in order to use those neural network and get hands-on.
Over a period of time in future lessons I will also show you how to build your own neural network from scratch.

Follow the below commands one at a time, each command / install will take 30-45 minutes depending upon your network.

Step 1 : Installing MatPlotLib, NumPy and PyPlot

sudo apt-get install python3-matplotlib

To confirm the installation, Open Python3 from terminal

import matplotlib.pyplot as plt
import numpy as np
np.__version__

Step 2 : Installing OpenCV

sudo apt-get install python3-opencv

Open python3 on terminal
import cv2
cv2.__version__

sudo apt-get remove python3-opencv

Installing Object Detection Inference Tools

Step 1 :

sudo apt-get install git cmake libpython3-dev python3-numpy
cd Downloads
git clone –recursive https://github.com/dusty-nv/jetson-inference
ls
cd jetson-inference
mkdir build
ls
cd build
cmake ../

Step 2 :
After executing all of the above commands one by one, you will get a screen showing up multiple pre-trained models for Object Detection and Image Recognition, select those models by pressing tab key and using down arrow key to enter ok.

Step 3:
After the models are downloaded you will get another screen showing PyTorch Installer. You will get 2 options are the bottom of that installer screen, use arrow key and select PyTorch 1.1.0 for Python 3.6.
Once PyTorch is configured, execute this command below within the same directory path as above – /Downloads/jetson-inference/build
make -j$(nproc)

Step 4:
Now you need to install the bundle within the same directory path ~/Downloads/jetson-inference/build$
sudo make install

Step 5:
Now you need to synchronize all the previous installers, tools, models etc, hence issue the below command
sudo ldconfig

Step 6:
Now you need to configure your USB Web camera parameter for which we need to setup a utility tool or you may call it a helper program
sudo apt-get install v4l-utils

Step 7:
Now you need to check if you properly installed PyTorch and TorchVision, in order to check that, you need to open Python on terminal :
python3
import torch
import torchvision

source