T3chFest 2019 (English)

Conference itself

T3chFest conference agenda:

Attended talks

Visual attention: How to guide a machine in the world around us, by Miguel Ángel Fernández Torres

He talked about using AI to see the world as human do, to help automate some tasks that are visually intensive.

What’s visual attention? recognize objects, track them, the human ability to focus on something outstanding.

Features: color, intensity, object orientation, movement, text, scale, shape,…

They use eye tracking as input data to know and learn how humans pay attention to what. So they need lots of this data.

With eye tracking data, they build maps (as heat maps), to create models:

Classic models (image processing,…) have evolved thanks to convolutional neural networks (or CNNs).

Input for these networks usually are color and movement.

Applications, what are they doing at the University:

  1. Object recognition: to help people with Alzheimer in their daily life and tasks
  2. Accion recognition, such as grabbing objects: to help impaired people
  3. Anomaly detection: to automate video surveillance

A framework-independent, decoupled frontend architecture for large-scale projects, by Carlos Villuendas

Carlos presented a front-end architecture, that allows them to build different sites with the same code base.

They structured the front-end architecture in tiers:

They started with 5 monoliths, for well-known Spanish sites: motor.es, fotocasa.es, coches.net, motos.net, milanuncios.es

At some point, they had a single JS file, with 13k lines of code, where jQuery was their architecture.

Knowledge must flow from outside to inside, from UI to business rules.

A place for everything, and everything in its place

Common components or theme define tokens to be overridden by specific components or themes.

The web app tier wraps and provides coherence to lower tiers. It doesn’t provide new features, it just wires things.

Compu-mathe-sci-tics, by Jennifer Wong

A talk about the Math object in JavaScript.

She started talking about history, from math to computer science.

She talked about how math and computer science understand the concept of function: for math it’s about input and output relationship, in CS, it’s about side effects and procedures.

Funny thing, the talk was on pi day (March 14th, 3.14)

ES6 adds lots of new methods to Math object: e**(x - 1), hypotenuse, trunc,…

JavaScript is not very good at precision. It’s admitted even in the standard. Most mathematical functions are approximations.

She could have demonstrated how most of the Math methods work, but the talk wasn’t too much interactive. It was a bit disappointing.

Bad UX, by Almudena Martín Castro

Talk about UX, users and graphic designs

There are too many apps, but too few designers.


  1. Provide control to the user, do not control them
  2. Feedback: there is no cause without effect, things happen because of something
  3. Minimize effort: mechanical (click), visual (noise, images, color) and cognitive (decisions, doubts)

Ten errors to fix:

  1. You’re not your user: users don’t know your product as well as you do
  2. There’s no such thing as a mean user: if you don’t measure, you don’t know what your users are using
  3. Users don’t read the text in your app: your app should be usable even if it’s blurred
  4. Users don’t have the latest devices (screens, light, internet connection,…)
  5. Users don’t take optimal decisions: they don’t think
  6. Do you think saving clicks is really important? clicks are cheaper than thinking
  7. Be consistent (similar elements do similar things), use conventions (extracted from other products)
  8. Your app must look nice
  9. Dark patterns: to cheat users, to lie them, to force them to do wrong things. Please, don’t do that, don’t lie to users.
  10. Arg, I missed one :(

I repeat a good idea:

clicks are cheaper than thinking

Let’s play with the open web, by Jesús M. González Barahona

The browser is the greatest platform ever. Much better than any operating system. It makes the old idea write once, run everywhere a reality, not Java.

Browser’s APIs are open, and we have different implementations. In fact, we need different implementations.

The speaker showed us and demonstrated us very interactive examples that are worth a try: Playing with the open web


  1. You can see your phone camera in the browser with 2 JavaScript lines
  2. Transform the input of your camera into ASCII art
  3. Maps with leaflet library
  4. Virtual reality, aframe library

With aframe you compose your VR scenes with HTML elements!! And you can manipulate the VR scene as a DOM document.

Really, really try some of the examples in your browser: Playing with the open web

Verifying fake news: Newtral, by Ana Pastor, Inés Calderón and Marilín Gonzalo

The speakers are 3 journalists, the main one is a well-known journalist for doing great questions in interviews. She has a TV program where some data provided by the interviewee is checked on real time.

Journalists must be where the public is, we must communicate in the same media as people communicate to each other (paper, tv, internet, messaging,…)

Newtral creates content that is easy to share, to help spreading the verification as fast as the news (fake or not).

Are fake news a issue for democracy?

Checking the facts is much harder than making up fake news. It so hard, and it takes so much effort that technology is needed to help journalist to do the job. They mention some machine learning terminology, but not too technical.

The problems with the data the speakers talk, remind me the problems people working with open data talk. Data is open, but it’s in different formats, it’s incomplete, with no too much coherence, out of date,…

Technology behind movies visual effects, by Carolina Jiménez García

I expected a technological talk, but it was more related about filming techniques. Anyway, it was quite visual and interesting.

It started with some history about cameras.

One main trick used for a long time is to use a glass, where half the scene has been painted, and film the movie scene through the glass, playing with perspective.

Then, there are some process, like Williams process or Dunning process, that lead us to the chroma.

First, it was used a blue chroma, because blue is the most different color from the human skin (making it easier to replace for other things during movie edition). Nowadays, in digital filming, a green chroma is used, because digital sensors are more sensitive to green than to blue or red.

Guardians of the Front galaxy, by Cristina Ponce and Cristina Fernández

They showed us lots and lots of suggestions for clean code (guarding us from bad code). The sad part is that they didn’t explain them deeply. At some point, they were just saying the name of the concept.

I got some nice ideas:

.editorconfig, it’s a convention (supported in PhpStorm with a plugin) to configure some parameters in software projects such as: tabs vs spaces, indentation,…

ESLint to check JavaScript style, best with published guides: airbnb, google,…

StyleLint, the same but for CSS

Husky, git hooks without the need of installing them on .git/hooks

Of course, they mention tests, all kind of them: unit, integration, end-to-end, coverage,… But, they didn’t go deep on any of them.

Team workflow: protect some branches, pull requests.

Automatic deployments. They showed a case with GitHub pages and Netlify. They use Circle CI, a tool similar to Jenkins.

Subscribe, like and another way to spread science, by Rocío Vidal, Santi García Cremades, Carlos DotCSV and Javier Santaolalla

The speakers are four well-known Spanish youtubers. One of them have a channel about AI and machine learning, and another one about math in general. The conversation was nice, because they’re used to interact with people. The idea was to have a moderator that asked some questions, and each of them give his own experience.

Are youtubers free about creating content? They say they’re more free than in a traditional media, but they say nothing whether they create content guided by likes or subscriber numbers.

All of them agree on the fact that they don’t want the tool, the mean, YouTube, to control what the want to communicate.

Can someone live out of YouTube? They agree on you can, but you never know for how long it will be that way. Maybe tomorrow you’re out of business.

Is YouTube adult? Other social media are much younger. Other social media are evolving to other activities (twitch -> games, instagram -> celebrities,…), but YouTube seems to have its place among all social media.

They told too that popular science has a great momentum right now.

They discussed about YouTube algorithm, and it seems to be as well known (or as well unknown) as Google search algorithm. The only thing they know is that it seems to like videos that keep the audience online, but it doesn’t mean the video is good. There are popular videos of irrelevant or even malicious things.

They were asked about why science audience seems to be mostly men. They talked about the algorithm here too. One cause can be that if YouTube knows you’re a man, it suggests you science channels. If you’re a woman, it suggests you other kind of channels.

Missed talks


I talked to a friend about push notifications. At work, they implemented offline notifications. They’re able to show notifications in the browser even when the tab of their product is closed. The tech behind this is Service Workers (and websockets for notifications I think). The idea is quite interesting: you login into EVSM, you close the tab, if anything new happens, you’re notified with a link to the task you need to perform. Nice!

Use leaflet maps instead of Google Maps. The library is open source, and it can show map tiles from different sources.

Este comic de CommitStrip viene que ni pintada para esto: