Merken

Postcards from GorbyPuff

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
thank and the things that we do and this is a set of my talk was postcard forgot about the other here have cited rather laugh something and other and prove that choice but I'm not and that was random and onto this talk is a man examples rebo already under rolls Conf 16 I treated that moves are really do people treated me during my talk my phone is somewhere over there so universe the mica interacting on talking and definitely will pop up along lines and my blog is Phenomizer . com besides the soccer also there were following a lot of and I really dinosaurs so I work for a
group of software and when I have a workshop tomorrow afternoon and workshops track if you want to carry stuff what sometimes it is worth 3 stuff of some voting and of embarrassing around that's great way to do all the out I wasn't good digital can after are questions get free stuff because that's why people have do this is to give out free stuff and if you were those people who doesn't like actually talking to people face-to-face and another some of the lack Ruby room several of those who work in the booth and simple to Wonderment me to the conference hang out that really we can answer questions that you know
because I worked for a big company I have this slide le Cancer as any John this talk is about regular license about you do owned by the way he did notice this is the cat pictures stopped there is probably an equal amount of captures and so I get so this is
this so it's going to be the he belongs
to this character who also lives in Seattle and is lovely is giving a keynote on Friday morning and eliminated only go to it because I currently k conditions less pictures of Gorbachev in this talk I stole them all from this conference
and where it was Madame couldn't come with general and rails golf so users on want tour of his very so in the red and the spirit of
19 eighties educational software when the world is going to be he's going to giving us closed form postcards you so that argument someone who gets the job that I am making here in the computer program that I am alluding to raise your hands the OK you that's all I care about this it and so I could this have a
statute they and use postcards views
about it God bless hanging out new stone friends tells me because of high and iron and so we can figure out what we had where's God read
and much like in that the quintessential eighties computer game we have access to a superset of problem-solving computer owners will
and were amusing who envision for us so that this is a Web API that can label landmarks faces face detection is employed there just only interface I texts they could also label large variety of things I'm going to go over get on and can always leave a large variety of things in images and it can also I find adult content of various sorts if that is the thing that is relevant to you you're not having more having for whatever your website might be but
I'm also going to be using energy content have you tried to using the as before the programmatically generated Germany had about time argument you come back and try you this is by reviews for it's hand-coded if you know Mike Moore or Chris Smith who were in Utah might use my grand the Mountain West x conferences and they're the ones who are working on this and fantastic job a real really feels natural so you take only need to look at the postcard from inquiry and walk through it 7 require the gem amusing visions of part of vision G-Cloud objects specifying 2 arguments here 1 is the project which is just a way of organizing your resources inside will cloud and the other is a key lets me anarchy into the project earlier this is for billing sadly vision is not free of vision is not free and a great vision objects and and image and when this image of a local desk and you can also call it from a web store of some sort or another and then this is the 1 that does all the magic some calling entity on the vision API passing at the images and telling it but I wanna find 1 landmark because really seem like there was only 1 in the picture and I something I'm afraid of the description is what's the most probable location which is going to be the lat long because yes from the picture can tell you what it is and where it is so because it's been a while that's what
the image looks like this is what comes out it actually
knows about the Fremont Troll from ontrol happens to be about 3 blocks from the Google Seattle office it using 10 things that about you you have seen this tall on it flies through the large part of memory so where the world is Gorbachev
Gorbachev is right here so you know a little like virtual a bit plane in the slide and we get this velocity around he says about the tagging is something for you Mike this get yeah yeah
that would be cap and he gave us this
postcard and the dimensions dimension to me that you know you want to hang out this creature he was very jealous of how not flat face for so here's the back of this 1
is a dog is a rat is the face and figure out what is it find it to read a little detection in group of vision this time and this can make all sorts of things like animals including breeze on dogs and cats especially if they're pretty pretty distinctive brains with the transportation types of structures in all sorts of other cool stuff and sneezing examples those are my cats was was taken
about a week ago a hanging out on the deck and I want to run this because visions you label it and so
these labels a given me and said the cats they're probably that's 1 them is kind of black actually breathing thing and black and that is the plastic cover on my Atlanta's on their mammals in that picture which is interesting the animals in that picture there probably playing their small-to-medium-sized cats and cat mechanic is tabby and so what cat like mammals you know was a little less or no matter what the user does that I'm not showing is that whenever it does any of these labels it gives you a presented certainty how confident is that a given a label applies it was a little less confident catlike mammal is another 1 entirely
different and this is the juridical skiing and these labels it came up with snow
mountain weather PC winter geological phenomenon and season an adventure constant Isabel white I'm was pretty happy with the label of adventure and 1 of things I want
to point out as you look at this picture of the figures now we think of not being white and is actually very very little light annotation because of lower lighting worked out but competition because the use of machine learning techniques is able to figure out this was now and the main thing is they think of snowmen cell it's critical of is that powerful barons 3rd example
on some bosses and says yet to
transport the devices not only on the bus the double decker busses probably into a bus service name commercial vehicles on and you know there's probably an engine and on the bus but it will be due on citizens who really really small amounts of cow specifically this amount of care at there's 7 lines including required so is basically the same setup as before the only thing that changes is this magic line instead of asking you to look for landmark and ask it to look for labels and Clinton using they are not fully release version library you can grab adopted habits and public repo and when they finally integrated fully into the Japanese animated so I have to use a and say it will find all the interesting stuff for and then because they don't have just a single landmark and then a linear rate over the labels that advances in print them out so back to the postcard we
had what the heck is so here I am pretty good guess missing some of you no idea what happens is didn't want that and that conversion
was successfully into that if I want that I tried it with a pangolin if you know pangolin as it thought it was a terminal on armadillo which is actually a fair amount of pangolin looks like so you know I'm not actually elements of that figure out what that pretty confidently of was 83 93 per cent confident there's was along that so generated
planning and fly past it's always
so that's what I thought that matching that description he told me nearly that's the the so here's another postcard from this is some
sort of sign some language that I don't currently speak on the back of it is tender love was
realistic countries he likes to visit meanings Lendl line around apartment just any idea what the signs that also began so I mentioned by the weakening text detection and the technical term for this is optical character recognition and what this does it'll find things the characters in the images and it'll figure out what character they are and the really cool thing is will also figure out what language they're likely and so the same set as before only minor changes instead looking for levels of for text and we print out the text analysis on a print out what but the idea locale is a combination of language in place so countries that have multiple official languages of multiple official cows countries that only have 1 official language that only have 1 look out and see what that it will still get that information we can figure out where he is gone after this time
so here's the picture again attacks the comes back it's not great it's not particularly accurate OCR is actually really hard especially often images on you have an image that is neatly lined up and using roman characters there's been a lot more training on those kinds of datasets and so tends to work better than using a non-Western characters are here and those lookout came back as j which is for Japan not surprising on tender lexical Japan he speaks Japanese not surprisingly gaudy would try to figure his owners been going off to all these times but actions speak Japanese full-sphere with us so users will translate that who's done trickery thing the real translate from 1 language to another to another internet planet the gradient of the 2 the fossil had so much and you never and there's support for translate into and this is I have to do to run something to translate making G-Cloud objects creating translate object from that and then running some translate text through I'm and point out that the from argument is completely and utterly optional it can usually gets pretty well based on the characters but since I have a low calendar thrown in because it will increase the likelihood of a translation as correct translate you know that it's doing its best human language is really really really really hard problem and so all getting really close and you generally get just sometimes in the lower house also this looking back from translate this that text and so you know there's a bunch of text in an image this isn't just outside is actually text on some of the buildings and stuff the background as well that couple then but I know that God is a flower garden and he's playing a beach in you cat giant sandbox this seems to make sense cell to catch with RBC when you do it that
plane someone says the cataloging forms here
he's been hitting the books yeah yes in thus far and stock and flow was going to Greece tumbler because you know he has a tumbler and Twitter in his work for the followers that idea so actually in my goal for this year official company goals of give more food to been were toward followers Gorbachev so
he's decided that he's going to give us some during this time what place on earth and the hottest temperature on actually May 4th there are 2 thousand and the slide didn't come out this is because God found that is not particularly good at really carry out the so he's going over really above trivia instead so the court told that Google has been created talk about this a amount West 2 years ago and it's a tool that we designed internally for processing large quantities of logs but the really cool thing I like about it is it it lets you query large datasets was equal as opposed to all those other databases or tools that use large datasets and you have to special grilling which 1 also it doesn't have any indices conflict is also magic trick you don't have to spend time fixing data just works in so cool so you want that we've got a bunch of public datasets while accusing is genes on which stands for global surface conditions of the day and into the database with the weather data it's a couple gigabytes you can do much bigger datasets I like this 1 because it's interesting everyone deals with whether and it's got a huge scheme including things like Boolean was the co-author on a given day but since we're trying to figure out what place in the entire world was hottest on May 4th 2000 where this is the stationary which didn't 5 locations the year month and day in the mean temperature for that makes some here the query so our many almost equal guess as long enough of it when I was doing a lot of interviewing write a sequel critical at approximately like this was pretty common 1 of my questions only here is in Table 5 minimax my walled off and this was a total acceptable answer not doing any additional of degree can totally do that it is a very quickly and defining the mean tampon ordered by descending so the top the 1st one's going to be the warmest the so the muscarinic having to do this is the demo afterwards monopoly touch my laptop the slides working right now and this is uh a won't where and less than 20 seconds ago the think 17 about 16 about about to do it on and it's cash so I run it again it's instantaneous after that and remember knows no indexes nothing like that just works as this is the 1st 4 rows that idea back in so this this place that was 102 degrees on May 4th development and the station number is this on ideas this dataset lot at this point and so I happen to know the leading 6 means it's in Africa and and this is a table that you can look up on the world theological Association wanted to figure out what these numbers are and what how happens with that this particular station was involved so let's go find ourselves again
yeah yeah need someone some shady
characters given as clear the cat had ball here
yeah I mean that 1 I spoke general of and good reason well almost
done the makes clear so the final pose God
looks like a landmark again someone integer out congratulations the rest of you will get shortly so I have been injured the worlds you're
about to be with me and but you know Goebbews Internet famous you flat ahead but spelling runs all things really boleros so the where he is so mark again same codes for the 1st compose terms out Coquin stadium right here in Kansas City the so congratulations we
found 1 about and also some of the
Swanson track and so as to make fun of myself because I not all my actual proposals rules come projected so I had be sponsored track and this talk was inspired by children's television from the eighties and I work for a company called alphabet so this talk today is brought to brothers G seed
and they that looks like a and so this is the sponsored pitch as sodium applications I promise you that we can most likely run them and the a lovely promissory there have run into anything that we can't run naming may not make some frustrated that up for you to decide how to talk to you as we've got a lovely team evoke some of those at the sectoral mocking me so there also have to help you out and we have the once we have storage we are great tools for big data that we have great tools for machine learning on a coming out all this stuff is available remain if you were like 1 of the like 5 people who actually use containers in production but was talking all some stuff for continued people of
and added it to make special announcement is I got to give the sponsored talk uh we're
announcing the Abidjan through is in the store as and today rection officially announcing it tomorrow because there's may the 4th Municipal start stores and establishes the battle but this is added is the platform as a Service project some products some of you may have tried a couple years ago were like we support happened really lonely these Jeremy that's not this product you actually use MRI like the use on your laptop will happily run it will happily all scale it will happily take care of learning a lot than if you put the right gems in you can create them with tools like the inquiry all that stuff were happy to do fourier on this product it's really super cool it launches well and it runs all the stuff you care about that a lot of people who worked with through for dozens of years of his life working on it and I'm really happy belly up here and say this and you answer a non-trivial 1 or else because some of the morals of materials confident I the other thing is you want to break the ice giants G-Cloud Mike and present on a fantastic job on this it's really much better experience was then using some the other parts of the past it just feels more like is the Fort create that might be now this was actually my romance was this feels like really like like the alias the heck out of things that entertainment has 5 times this using out it's really means do everything at least 3 different ways and then all the right we're not quite so this is something you are
Harrison if I give the idea of this talk at work in the light of the enacted deterrence permission these all the pictures of his cat is because the best idea ever will know that he is not in this talk so of his actions in the slides already a might more on Christmas for the decoder and my lovely co-workers and other folks who could make a worked on G called Ruby with me the group under the support I have been working in Google for about 18 months and finally landing to talk about something Yale can use it is a doctor makes me happy so I think that's
all creativecommons by 2 . 0 a this followed the questions and I am repeating
something as the earlier this year if you ask a question also you can so yes I like to use the platform adopted why not ball the ball and so if you don't wanna do the office work yourself trout happens and if you want if you like kind of a set of VM try compute engine if you like his Dr. things scenes are so you can try out of consideration we have stuff to support all 3 levels of recall the compute continuum because like big words like that so but we have a very much have something for everyone and no place where delegates to be an interesting conversation is if a particular competitor or running stuff on premises so it is more effective for you know not succumb prices keep dropping so you know checkers on a couple months it may be cheaper than of so capuchin this 1 is then
informed and you add random useless facts when you're well known Rubia's fosters baby kittens and I get to take pictures of them yes Mac I did not do other commonly claimed that so if you need a something from home to the question was what did I do the the inquiry great Jakarta they do it some other way so there's 1 interface actually like the web interface because I can save things and export them but to various formats and it just how many you guys have built an ad-hoc reporting tool for a company you've worked at yes OK big big like can be that reporting tool is for you that's 1 lovely things about it is that the weather face you can change your mind about queries and explore and do things like that I know folks were using it from the command from the GP gender from code because they're doing queries over and over again and they want them to show on unattached words and so don't don't doing it from G call because they know what degree they want most of the mandatory exploratory stuff so I use the light but it works it's just as fast it's got a pretty I love you said margins this 1
was and that's arranges orange given some often often it's any other questions yes get so what I was reminded vision software I don't actually know intended it I know there are a number of good and the idea is not that's all I know the sum the folks who worked on it also worked on tensor flow which is our open source machine learning library that you can download and try use for all of your various machine learning problems and they don't actually the time of the details of the or some details available online and I know that people are working on white papers and deeper talks on the actual machine learning part behind it where things I found is that like I'm a giant pile of data and a lot of this really cool stuff what do you like and relevant to machine learning books and took the Coursera Machine Learning class but unlike I could start my timer implement this and it'll grab something that someone else has done and do the cost-benefit analysis on July think things like the television and we've actually got was the talk you will and it's been announced no Speech API that is a new 1 for speech as well so which I've encode a lot of very weird cool things to do with that it did you could just send an audio file and give the text that can also from a true for an arbitrary corpus of of words than pelicans so as
not but the thing was was assumed announced you got on is that what you get is a coming soon there to these 2 people working on a full-time it's completely open source it's up on you have you can't always look at it you can enter issues and that's how I got that vision so might have a talk on this ingeniously that vision and nicely yes let vision and it's not gonna be fully integrated by the time needed like OK that's of some working on action working of a branch of that if you look at my code on get have under my report you'll see you the actual and Gentile C. 1 point from also so remain out you want to us I also have a the cardboard if you want to see a 3 D 2 were like 1 of our data centers you can totally come over and show the head face up against halocarbons 3 2 or 1 reasoners and unlike in the data center has blinking lights but it turns out really being dead so as they're actually actually really cool even if you think that would you like to boring so they got thank you thank you and the of
Menge
Web log
COM
Gleitendes Mittel
Grundraum
Auswahlaxiom
Gerade
Computeranimation
Metropolitan area network
Motion Capturing
Rechenschieber
SCI <Informatik>
Abstimmung <Frequenz>
Differenzkern
Freeware
Software
Code
Gruppenkeim
Ablöseblase
Computerunterstützte Übersetzung
Computeranimation
Konditionszahl
Parametersystem
Bildschirmmaske
Prozess <Informatik>
Besprechung/Interview
Optimierung
Computeranimation
Lernprogramm
Teilmenge
Computerspiel
Sichtenkonzept
Grundsätze ordnungsmäßiger Datenverarbeitung
Computer
Gesetz <Mathematik>
Computeranimation
Streuungsdiagramm
Parametersystem
Web Site
Quick-Sort
Computeranimation
Eins
Objekt <Kategorie>
Spezialrechner
Energiedichte
Deskriptive Statistik
Benutzerbeteiligung
Prozess <Informatik>
Mereologie
Projektive Ebene
Maschinelles Sehen
Inhalt <Mathematik>
URL
Speicher <Informatik>
Maschinelles Sehen
Bildgebendes Verfahren
Varietät <Mathematik>
Festspeicher
Mereologie
p-Block
Bildgebendes Verfahren
Computeranimation
Office-Paket
Rechenschieber
Ebene
Geschwindigkeit
Bit
Hausdorff-Dimension
Computeranimation
Kugelkappe
Sichtbarkeitsverfahren
Datentyp
Gruppenkeim
Datenstruktur
Computerunterstützte Übersetzung
Transportproblem
Figurierte Zahl
Maschinelles Sehen
Quick-Sort
Computeranimation
Kraftfahrzeugmechatroniker
Bereichsschätzung
Computerunterstützte Übersetzung
Computeranimation
Überlagerung <Mathematik>
Konstante
Virtuelle Maschine
Abenteuerspiel
Zellularer Automat
Figurierte Zahl
Drei
Abenteuerspiel
Computeranimation
Leistung <Physik>
Hochdruck
Versionsverwaltung
Dienst <Informatik>
Bitrate
Computeranimation
Spezialrechner
Dienst <Informatik>
Mehrrechnersystem
Benutzerschnittstellenverwaltungssystem
ATM
Programmbibliothek
Bus <Informatik>
Delisches Problem
Maschinelles Sehen
Repository <Informatik>
Gerade
Umsetzung <Informatik>
Bereichsschätzung
Hecke-Operator
Radikal <Mathematik>
Element <Mathematik>
Bitrate
Figurierte Zahl
Computeranimation
Deskriptive Statistik
Vorzeichen <Mathematik>
Formale Sprache
Automatische Handlungsplanung
Quick-Sort
Computeranimation
Wellenpaket
Formale Sprache
Hochdruck
Mathematisierung
Schaltnetz
Gruppenoperation
Natürliche Zahl
Zellularer Automat
Term
Computeranimation
Gradient
Internetworking
Übergang
Multiplikation
Mehrrechnersystem
Vorzeichen <Mathematik>
Schriftzeichenerkennung
Figurierte Zahl
Gerade
Bildgebendes Verfahren
Analysis
Parametersystem
Likelihood-Funktion
Gebäude <Mathematik>
Stellenring
Natürliche Sprache
Objekt <Kategorie>
Arithmetisches Mittel
Menge
Grundsätze ordnungsmäßiger Datenverarbeitung
Information
Computerunterstützte Übersetzung
Chinesischer Restsatz
Ebene
Bildschirmmaske
Twitter <Softwareplattform>
Datenfluss
Computeranimation
Maschinenschreiben
Demo <Programm>
Flächentheorie
Versionsverwaltung
Zahlenbereich
Login
Computeranimation
Eins
Datensatz
Vier
Flächentheorie
Notebook-Computer
Arbeitsplatzcomputer
Statistische Analyse
Konditionszahl
Indexberechnung
Ordnung <Mathematik>
Softwareentwickler
Datenhaltung
Zwei
sinc-Funktion
Abfrage
Nummerung
Arbeitsplatzcomputer
Rechenschieber
Minimalgrad
Automatische Indexierung
Konditionszahl
Boolesche Algebra
URL
Mini-Disc
Tabelle <Informatik>
Ganze Zahl
Grundsätze ordnungsmäßiger Datenverarbeitung
Computerunterstützte Übersetzung
Computeranimation
Spezialrechner
Rechter Winkel
Codierung
Maschinelles Sehen
Term
Computeranimation
Internetworking
Weg <Topologie>
Familie <Mathematik>
Zeichenvorrat
Schlussregel
Computeranimation
Streuungsdiagramm
Virtuelle Maschine
Systemplattform
Kartesische Koordinaten
Speicher <Informatik>
Biprodukt
Analytische Fortsetzung
Computeranimation
Streuungsdiagramm
Videospiel
Zentrische Streckung
Aliasing
Materialisation <Physik>
Hecke-Operator
Gruppenoperation
Gruppenkeim
Biprodukt
Kombinatorische Gruppentheorie
Systemplattform
Computeranimation
Rechenschieber
Vier
Prozess <Informatik>
Betafunktion
Notebook-Computer
Mereologie
Projektive Ebene
Computerunterstützte Übersetzung
Speicher <Informatik>
Demoszene <Programmierung>
Umsetzung <Informatik>
Menge
Kontinuumshypothese
Wort <Informatik>
Technische Informatik
Systemplattform
Computeranimation
Digitale Photographie
Übergang
Office-Paket
Randverteilung
Gewichtete Summe
Benutzeroberfläche
Open Source
Güte der Anpassung
Klasse <Mathematik>
Systemaufruf
Abfrage
Sprachsynthese
Datenfluss
Code
Virtuelle Maschine
Minimalgrad
Tensor
Software
Geschlecht <Mathematik>
Mereologie
Programmbibliothek
Dateiformat
Wort <Informatik>
Algorithmische Lerntheorie
Maschinelles Sehen
Verkehrsinformation
Analysis
Schnittstelle
Rechenzentrum
Punkt
Open Source
Gruppenoperation
Maschinelles Sehen
Schreib-Lese-Kopf

Metadaten

Formale Metadaten

Titel Postcards from GorbyPuff
Serientitel RailsConf 2016
Teil 80
Anzahl der Teile 89
Autor Hammerly, Aja
Lizenz CC-Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Unported:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
DOI 10.5446/31542
Herausgeber Confreaks, LLC
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract GorbyPuff is setting off on a grand world tour and leaving us clues so we can follow him. Using Google Cloud tools such as Cloud Vision, BigQuery, Google Translate we'll track down Gorby at all his stops. While we're at it we'll learn some tools you can use to add awesome functionality to your web applications.

Ähnliche Filme

Loading...