Open Source TensorFlow Models (Google I/O '17)

Come to this talk for a tour of the latest open source TensorFlow models for Image Classification, Natural Language Processing, and Computer Generated Artwork. Along the way, Josh Gordon will share thoughts on Deep Learning, open source research, and educational resources you can use to learn more.
See all the talks from Google I/O '17 here: goo.gl/D0D4VE
Subscribe to the Google Developers channel: goo.gl/mQyv5L
Follow Josh on Twitter: / random_forests
#io17 #GoogleIO #GoogleIO2017 event: Google I/O 2017; re_ty: Publish;

[@GarethHall]

See Josh in thumbnail, immediately click video.

[@doberek]

same here :)

[@nishatsayyed8326]

his smile while he is speaking is so engaging....... appreciated🔥👍👍

[@GarethHall]

Thank you Josh, awesome talk. How can you not like this guy?

[@chenmaus]

great content and great presentation, google just have so many amazing people that could keep the company great

[@khrumkashan6954]

Thank you Josh ! Such a awesome guy and great video loved it .

[@zhangtalen6130]

can you share all model names and their open source links, much appreciated

[@JBGordon]

Sure thing! In fact, here are all the links from the talk, starting with the open source models.
The three open source models I introduce:
* research.googleblog.com/2016/08/improving-inception-and-image.html
* research.googleblog.com/2017/03/an-upgrade-to-syntaxnet-new-models-and.html
* research.googleblog.com/2016/10/supercharging-style-transfer.html
Medical applications using Inception:
* research.googleblog.com/2016/11/deep-learning-for-detection-of-diabetic.html
* news.stanford.edu/2017/01/25/artificial-intelligence-used-identify-skin-cancer
* research.googleblog.com/2017/03/assisting-pathologists-in-detecting.html
More shared research in TensorFlow:
* magenta.tensorflow.org
* research.googleblog.com/2017/04/introducing-tf-seq2seq-open-source.html
* research.googleblog.com/2016/05/announcing-syntaxnet-worlds-most.html
* research.googleblog.com/2016/09/show-and-tell-image-captioning-open.html
* research.googleblog.com/2016/08/text-summarization-with-tensorflow.html
Deep Dream notebook and other code examples:
* goo.gl/TjQPfS
Michael Nielsen’s book
* neuralnetworksanddeeplearning.com
TensorFlow for Poets 1 & 2:
* goo.gl/xGsB9d
* goo.gl/9NbWyw
Android samples
* goo.gl/AFqm5E
Chris Olah's blog:
* colah.github.io
Learning more:
* cs20si.stanford.edu
* cs231n.stanford.edu
* goo.gl/KewA03

[@ferencschultesz6389]

Amazing, thank you for this!

[@walterwhite5249]

Josh Gordon plz talk with TH-cam team and modify the like button so that we can like multiple times not only one to show our gratitude towards your presentation it was fabulous

[@kennedynganga4210]

@13:21 is when he touches on the most interesting part of his whole talk>> Transfer Learning. Wish he'd gone into more detail and practice on this.

[@lupusk9productions]

omg screw watch repair i need to go into this stuff. So exciting!!!

[@lightowl4345]

@ approx 23 minutes he describes something that is done by toolwiz photos, one of my favorite apps.

[@JasonMayes]

Can the multibox detector be retrained to recognise and locate custom objects eg cars not people? If so, link to tutorial for that (just the TF part, no need for Android part).

[@JBGordon]

Yup. I was just exploring YOLO2 (hilarious name) that detects a number of objects out of the box. You can try it here: github.com/thtrieu/darkflow

[@JasonMayes]

Thanks Josh :-) Shall take a look

[@JOKBO1]

Such a great presentation!

[@jcjensenllc]

either scale the talk to to the alotted time or schedule more time. I want to see slides when mentioned, I don't need to see speaker.

[@vibhanshusharma3150]

Big Fan....

[@chrissharkeyai]

Fantastic talk!

[@muhammadhannan6701]

Tensorflow re-trained model on Inception-V3 takes longer to predict than the Inceptio-V3 provided. How can we optimize our model?

[@GughaGSrinivasan]

Hot dog or not a hot dog.... can we use tensor flow for that? :P

[@mertburan]

it is already used for that: news.ycombinator.com/item?id=14347467

[@JBGordon]

Actually, yes! You can build your own hot dog / not hot dog Android app using TensorFlow for Poets #1 (to train the model) and TensorFlow for Poets #2 (to install it on Android):
* codelabs.developers.google.com/codelabs/tensorflow-for-poets
* codelabs.developers.google.com/codelabs/tensorflow-for-poets-2

[@yixunnnn]

Hahah Jian Yang!!!

[@jobiej7416]

Hi. How can I use tensorflow in a progressive web app.

[@CommonOnFire]

This should be a really big deal.
But I found something in github.com/johannest/ai-experiments

[@lilaiwan3309]

a great company

[@athuldevin]

i have a plan to develop a generative chat bot using tensor flow , from where should i start

[@siddharthmittal9355]

lol same

[@tylerbarker4763]

from eye to cone , I just took your aligorthym/algorithm name to sound city populationTGB - Tyler Guy Barker

[@tylerbarker4763]

by definitionTGB real life word/words break words to life /TGB

[@tylerbarker4763]

googlemania ropeTGB

[@walterwhite5249]

I think all those who have been disliked should be & can be classified as *fools* By using *Tensor flow classifiers*....

[@EscapeAdventures]

Work on dementia and alzheimers prevention :)

[@diegushio91]

no real dogs where caged during this video.

[@jayel444]

🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋🙋HELLO WORLD