CyberPunk ˈsaɪ.bə.pʌŋk

Original Article Published 4th April 2018

USING PYTHON CODE, TENSORFLOW, C++, LINUX COMMANDS

From Urban Dictionary

The more recent lifestyle that has evolved and coined this term is fragmented and hard to classify. It is still very new and there is debate on what the term means. Generally observation shows that "cyberpunks" are not only extremely technologically proficient, but are innately adept at it. They have a desire to know all forms of technology they come into contact with. 

If you have a desire to learn, you also must not ignore a desire to have fun. 

Here we learn how to connect and stream fast wireless video from a Raspberry Pi to a Ubuntu Laptop.

Pi Hardware

Step 1) Install Raspbian stretch With Desktop onto the Raspberry pi (2/3/Zero W) here.

Step 2) Connect and enable the Pi Camera module here.

Step 3) Open a terminal and find out your ip address on the Pi once connected to wifi here.

Step 4) put all in a nice case like this one

(Pi Zero W)

Ubuntu Laptop

Step 1) Install Ubuntu 16.04 onto a hard drive here.

Step 2) Open a terminal and find out your ip address on the Laptop once connected to wifi here.

Stream Video

Step 1) Open Putty on the laptop and connect to the Rpi using the ip here.

Step 2) Copy and paste this line of code into putty and send it to the pi to begin streaming:

raspivid -t 0 -w 640 -h 480 -fps 30 -b 1700000 -o - | gst-launch-1.0 -v fdsrc ! h264parse config-interval=1 ! rtph264pay ! udpsink host = (your ip here) port= 9000

Step 3) Open Terminal on the Laptop and copy and paste this into:

udpsrc port=9000 ! application/x-rtp,encoding-name=H264,payload=96 ! rtph264depay ! avdec_h264 ! videoconvert

Watch the Video on the Laptop

Tips:

-Both Raspbian Stretch & Ubuntu 16.04 have gstreamer installed already

-You can increase the video frame size up to 1080p

-You can rotate the pi camera video using code -rot 180 


CRITICS I CAN HEAR YOU "Where's The Ai!?"

Well....

Here we learn how to pipe the video stream into a Movidius Neural compute stick and perform Ai experiments.

Laptop

Step 1) Install the Movidius SDK using this simple get started page here.

Step 2) Clone the Git repo here to a folder on your laptop

Step 3) 'Make all' the examples

Step 4) Play with Examples

Once you get to know your way around the examples in the NCAPPZOO Git, you can start to understand more.

With a little bit of knowhow you can stream the UDP pi video into the Neural Compute Stick for processing using just some Python code.

To do this you must be able to write and edit python script.

Here are some examples I made using facenets as Ai platform. You can use it to identify Gender and Age of people, or if they are Happy or Sad.

Amazon are using something similar with their Rekognition product, they are probably using facenet too.

New & Interesting:
News Article from The Independent - 'Amazon Shareholders demand company stop selling facial recognition technology to governments' (June 2018)

Amazon's response here

Python examples

I was on the vodka last night and feeling controversial. How can we use face categorisation? What might be a target market? Estate Agents? Property Developments? Who might not want certain demographics to live in their new developments? It's happening people, believe me. So with a few shots of vodka, and some courage, I made these examples

1) Feminist/Not Feminist

2) Jewish/Not Jewish

3) Islamist/Not Islamist 

Names Not Down You're Not Coming In...
Remember that music video The Bouncer?

Robot Ai door policy on it's way. I promise.

Have fun. Don't take life too serious, there are many many things to experiment with using Ai, and lots not yet even realised. It is one of the great things about a new emerging technology. Remember Cyberpunk rule #101 you are just showing the world what already exists out there. Don't be offended.

Tune in next week as I down a 5th of Vodka, mount all this to a drone and use Ai to command auto-attack. 👍

Update June 2018:
Now also being discussed (ethics) in a Register Article here

Google axes Maven for drone analysis here 

more here