A couple of months ago, I attended the Google Cloud Next 17 event in London. Among the talks about SRE, and keynotes, I had the chance to listen to Martin Gorner’s excellent introduction: TensorFlow and Deep Learning without a PhD, Part 2. If you don’t want to look at the video, here is a quick summary:
a 100 of lines of python are reading all Shakespeare’s plays; it learns his style, and then generates a brand new play from scratch.
In a previous post, I did some experiments with gRPC, protocol buffer and Terraform. The idea was to transform the “Terraform” cli tool into a micro-service thanks to gRPC.
This post is the second part of the experiment. I will go deeper in the code and see if it is possible to create a brand new utility, without hacking Terraform. The idea is to import some packages that compose the binary and create my own service based on gRPC.
This post is a little different from the last ones. As usual, the introduction tries to be open, but it quickly goes deeper into a go implementation. Some explanations may be tricky from time to times and therefore not very clear. As usual, do not hesitate to send me any comment via this blog or via twitter @owulveryck.
TL;DR: This is a step-by-step example that turns a golang cli utility into a webservice powered by gRPC and protobuf. The code can be found on GitHub.
Imagine a CCTV at home that would trigger an alert when it detects a movement.
Ok, this is easy.
Imagine a CCTV that would trigger an alert when it detects a human (and not the cat).
A little bit trickier.
Now imagine a CCTV that would trigger an alert when it sees someone who is not from the family…
Disclaimer: This article will not cover everything. I may post a second article later (or not). As you may now, I am doing those experiments during the night as all of this is not part of my job. I hope that I will find time to actually conclude the experiment. If you are a geek and you want to test that with me, feel free and welcome to contact me via the comments or via twitter @owulveryck. In this article, I will describe the method. I will also retrain a neural network to detect people. I will also use a GO static binary to run it live and evaluate the performances. By the end, I will try a static cross compilation to run it on a raspberry pi, but as my rpi is by now out-of-order, I will test it on qemu.
Google has recently announced its new service called “Google Cloud Video Intelligence”. The purpose of this service is to offer tagging and annotations of digital videos.
I will try this service on a trailer of a French parody. This movie is made of several scenes taken from erotic movies of the seventies.
Why this parody?
Caution: this video is not a porn video, but is indeed not safe for work (#nsfw)
TL;DR: Thank you for passing by. This article is, as usual, geek oriented. However, if you are not a geek, and/or you are in a hurry, you can jump to the conclusion: Any real application?
During the month of may, I have had the chance to attend to the Google Next event in London and the dotAI in Paris. In both conferences I learned a lot about machine learning.
What those great speakers have taught me is that you should not reinvent the wheel in AI. Actually a lot of research is done and there are very good implementation of the latest efficient algorithm.
I have discovered the notion of extreme programming more than 15 years ago. My job was to integrate and to develop pieces of code in Java for the IBM Websphere Business Integration server. We were a small team with light programming skills. A part of our job was to operate the software, the other part was to develop. It was in 2003.
We were trying hard to stick to the specific framework we developed.