What are some fun things to do with Keras?

what are some fun things to do with Keras?

Attached: keras-logo-2018-large-1200.png (1200x348, 49K)

Other urls found in this thread:

elitedatascience.com/machine-learning-projects-for-beginners
twitter.com/NSFWRedditImage

I'm a NN noob so anything really. MNIST can only get so exciting

whatcha wanna know young man?

I dunno what I want to know

Detecting emotion in audio or images etc. is a very interesting topic of research. I'm trying to do a thesis in it (but failing though)

Good luck

have you done these?

elitedatascience.com/machine-learning-projects-for-beginners

How about making a sarcasm detector?

Attached: Sarcasm_Detector.png (350x263, 116K)

don't make autists invincible

NN noob here trying to build an age classifier by fine tuning VGG16. Validation loss keeps increasing even with data augmentation and dropout. Any tips?

Lower your learning rate.

make an image to text generator

Try with GANs. Some 16 y/o kid I followed on Twitter made the news by using GANs + scraped images from a fashion designer website.

Not OP but how hard would it be to make a program that listen the input from a microphone, triggers when there's a baby cry, and maybe classifies the cries to know if the baby is hungry, shat itself, wants to burp and so on?

shitpost RNN

Keras main dev is hysterical uber shitlib who sees nazis and fascism everywhere and blocks anyone who isn't equally hysterical instantly on Twitter. Check his feed, it's fucking awful.

uninstall and install pytorch

Do you have a baby to collect data?

This. Not only is Pytorch much better designed, its also not controlled by rabid SJWs.

Not yet but considering my pace on side projects, I better get started soon :^)
There are data available online but the quality is quite rubbish.
Anyway, instead of going straight "neural net lmao", I was wondering if there are other more efficient approaches for this specific kind of problem.

>expecting Jow Forums brainlets to know anything about non-consumertech let alone something as advanced as deep learning
u wot mate

Attached: 1526972046743.jpg (289x440, 20K)

It wouldn't be that hard if you had tons of classified data already.

now i wanna to make a (((coincidence))) detector CNN with keras