#مقاله
تشخیص چهره جعلی
FaceForensics++: Learning to Detect Manipulated Facial Images
[pic: https://www.groundai.com/media/arxiv_projects/504837/img/teaser_bg.png ]


https://arxiv.org/abs/1901.08971

———————————-
مرتبط با:

virtual_talking تبدیل حالات چهره توسط یادگیری عمیق:
https://news.1rj.ru/str/cvision/349

خبر جعلی از اوباما
https://news.1rj.ru/str/cvision/532
Face2Face
https://news.1rj.ru/str/cvision/231
You said that? - Synthesizing videos of talking faces from audio
https://news.1rj.ru/str/cvision/230

#آموزش

در مورد پست مدیومی که امروز در اینجا از تنسرفلو 2 منتشر کردم، و مباحثی که در اینجا مطرح شد. François Chollet در باز توئیت به پست تنسرفلو مجموعه مباحثی را در اینجا مطرح کرده که خواندن آنها خالی از لطف نیست.

Symbolic APIs are APIs to build graphs of layers. Their strong points are that: - They match how we think about our networks (NNs are always visualized as graphs of layers in textbooks & papers) - They run extensive static checks during model construction, like a compiler would

This gives you the guarantee that any model that you can build, will run. The only form of debugging you'd have to do at runtime would be convergence-related. The UX of these APIs is highly intuitive and productive

Meanwhile, the subclassing API has the look and feel of objected-oriented Numpy development. It's ideal if you're doing anything that cannot easily be expressed as a graph of layers, and you feel comfortable with software engineering best practices and large Python projects.

It will involve execution-time debugging, more code, and will expose a greater error surface, but at the same time it will give you greater flexibility to express unconventional architectures.

Importantly, in TF 2.0, both of these styles are available and are fully interoperable. You can mix and match models defined with either style. At the end of the day, everything is a Model! That way, you are free to pick the most appropriate API for the task at hand.

In general I expect ~90-95% of use cases to be covered by the Functional API. The Model subclassing API targets deep learning researchers specifically (about 5% of use cases).

I think it's great that we don't silo researchers and everyone else into completely separate frameworks. It's all one API, that enables a spectrum of workflows, from really easy (Sequential) to advanced (Functional) to fully flexible and hackable (Model subclassing)

#tensorflow #keras #symbolic #imperative

#خبر #fastai

اخیرا کورس نسخه 3 fastAI منتشر شد (اینجا) و دیدیم که گوگل کولب هم پشتیبانی این فریم ورک را اضافه کرد (اینجا) اکنون میتوانید تمامی نوت بوکهای این کورس را به عنوان Kaggle kernel استفاده کنید.

you can now run all the whole http://course.fast.ai lessons for free using Kaggle kernels.

https://course.fast.ai/start_kaggle.html

#آموزش #کولب
Get started with Google Colabs (Coding TensorFlow)
https://www.youtube.com/watch?v=inN8seMm7UI&list=PLQY2H8rRoyvwLbzbnKJ59NkZvQAW9wLbx&index=2

the first episode of the Intro to Google Colaboratory series

including code and text cells, data visualization, sharing notebooks, and more!

#منبع #کورس
اتومبیل های خودران یکی از مهمترین برنامه های کاربردی هوش مصنوعی در جهان خواهد بود.اگر به این موضوع علاقه دارید در سایت coursera، یک دوره توسط دانشگاه Toronto شامل کورس های زیر اضافه شده است:

1- Introduction to Self-Driving Cars
2- State Estimation and Localization for Self-Driving Cars
3- Visual Perception for Self-Driving Cars
4- Motion Planning for Self-Driving Cars

https://www.coursera.org/specializations/self-driving-cars

#Self_Driving_Cars

اصلاح خطای توالی ژنی با تنسرفلو
Using Nucleus and TensorFlow for DNA Sequencing Error Correction

blog post:
https://medium.com/tensorflow/using-nucleus-and-tensorflow-for-dna-sequencing-error-correction-47f3f7fc1a50

gogle colab:
https://colab.research.google.com/github/google/nucleus/blob/master/nucleus/examples/dna_sequencing_error_correction.ipynb

#DNA_Sequencing

#خبر
استفاده از Keras برای شناسایی خودکار supernovaها که زمان را به طور چشمگیری نسبت به زمانی که یک ستاره شناس باید صرف کند کاهش داده و با دوبرار سرعت انجام میشود...

How 3 engineers built a record-breaking supernova identification system with deep learning

https://medium.com/@dessa_/space-2-vec-fd900f5566


#keras #supernova

#خبر #مجموعه_داده

اولین #دیتاست دیوار ریلیز شد.
این دیتاست شامل حدوداً یک میلیون پست در سایت دیوار است.
Published on 2019/01/30


https://research.cafebazaar.ir/visage/datasets/

#سورس_کد استفاده ساده از GAN در پایتورچ
Open-source VeGANs, a small library to easily train various existing #GANs using #PyTorch.
You provide a generator and discriminator, and VeGANs trains them with a GAN algorithm of your choice.
https://github.com/unit8co/vegans #pytorch #GAN #MachineLearning

استفاده ساده تر از GAN ها در تنسرفلو و کراس
GAN-Sandbox
has a number of popular GAN architectures implemented in Python using the Keras library and a TensorFlow back-end.


Vanilla GAN implemented on top of keras/tensorflow enabling rapid experimentation & research. Branches correspond to implementations of stable GAN variations (i.e. ACGan, InfoGAN) and other promising variations of GANs like conditional and Wasserstein.


Standard GAN implemented on top of keras/tensorflow enabling rapid experimentation & research. Branches correspond to stable implementations of GAN architectures (i.e. ACGan, InfoGAN, Improved wGAN) and other promising variations of GANs (i.e. GAN hacks, local adversarial loss, etc...).

https://github.com/mjdietzx/GAN-Sandbox

#GAN #keras

چرا آموزش شبکه های GAN دشوار است؟
#GAN — Why it is so hard to train Generative Adversarial Networks!

https://medium.com/@jonathan_hui/gan-why-it-is-so-hard-to-train-generative-advisory-networks-819a86b3750b

استفاده از image captioning برای پیدا کردن تصاویر مشابه در فضای بردار ویژگی fastText

Facebook BISON: An Alternative Evaluation Task for Visual-Grounding Systems
https://medium.com/syncedreview/facebook-bison-an-alternative-evaluation-task-for-visual-grounding-systems-857bb5366487

#خبر
https://www.wired.com/story/worlds-fastest-supercomputer-breaks-ai-record/

🙏Thanks to: @MohsenF91

#gpu #سیستم

#مجموعه_داده

Hotels-50K: A Global Hotel Recognition Dataset

یک دیتاست خاص!
یک میلیون تصویر از ۵۰۰۰۰ هتل در سراسر جهان.
نکته جالب: اینکه طرف می‌خواد ازش برای شناسایی هتل‌هایی که توش انسان قاچاق شده استفاده کنه.
نکتهٔ بامزه: توییت بعدی طرف گفته «با وجود عکس‌های اتاق خواب، این مقاله GAN نیست» :))

مقاله:
https://arxiv.org/abs/1901.11397

گیت هاب و داده:
https://github.com/GWUvision/Hotels-50K

لینک به توئیت مرتبط:
https://twitter.com/dennybritz/status/1091312057973014528

🙏Thanks to: @samehraboon

#dataset

#معرفی_سایت
Browse state-of-the-art
https://paperswithcode.com/sota

🙏Thanks to: @vahidreza01

Facial Recognition and Modelling subtasks
https://paperswithcode.com/area/cv/face

more topic:
https://news.1rj.ru/str/cvision/976
#face

#خبر
در این خبر تعداد 27,000 پردازنده گرافیکی یا GPU در آزمایشگاه ملی Oak Ridge برای اجرای مدل تنسرفلو جهت تشخیص الگوهای آب و هوایی و پیش بینی آینده استفاده شده است...

#tensorflow

#مقاله

The Evolved Transformer
The Evolved Transformer: They perform architecture search on Transformer's stackable cells for seq2seq tasks. “A much smaller, mobile-friendly, Evolved Transformer with only ~7M parameters outperforms the original Transformer by 0.7 BLEU on WMT14 EN-DE.”
https://arxiv.org/abs/1901.11117

The Evolved Transformer is twice as efficient as the Transformer in FLOPS without loss in quality.

#seq2seq

مدل های pretrained و عنوان مقالاتشان

🙏Thanks to: @AM_Ghoreyshi

#pretrained

#caffe #tensorflow #keras #pytorch