reiinakano

Ahh I see, these are some resources that should help you.

The original Neural Style Transfer paper should help you understand the loss functions involved: https://arxiv.org/abs/1508.06576

The paper introducing Real-Time Style Transfer is basically the algorithm used here: https://arxiv.org/abs/1603.08155

I guess this is Github's issue then? Things should settle down in a few days, it might just be having trouble with the HN traffic.

I'm not very well versed with the web, what does 'bundle.js?<anything>' do? Will adding that to my code solve the problem?

You can see one result in the repository README! :)

Maybe clear the cache? Sorry not sure how to debug as it's up for me on all my browsers + curl

Fixed now. Sorry!

Fixed now. Sorry!

Things are currently quite inefficient as there are some non-optimized (non-WebGL) calls being made. DeeplearnJS is a new library but once optimized functions are available, I'll update the site. :)

It is currently VERY slow since there are a lot of non-optimized (non-GPU) calls being made. Once those are implemented in deeplearn.JS, things should run much faster. In fact, one of the creators of deeplearn.JS says it might be feasible to do this realtime in a webcam (although not at 60FPS).

This was a super quick PoC demo I hacked together in a week on a very new library. :)

Fast Neural Style Transfer works like this: Each style image is actually trained on a neural network outside of the browser (using Python on Tensorflow) for around 6+ hours with a GPU.

The output is a neural network that can take any content image and outputs it in the style it was trained on.

The browser does this: 1. Downloads the model (6.6MB~) for the particular style. 2. Does one pass through the content image and outputs the styled image.

No training is done in the browser whatsoever.

Works now. Sorry guys. I tried to hotfix something into production and screwed things up. This is a lesson I will not forget... Lol.

"Upload from file" option not appearing?

DeeplearnJS does not work on mobile at all unfortunately. It's a new library, they'll get there. :)

Updated!

There should no longer be a problem with Twitter. The way I saved the styled photo was Right click > Save image in Chrome.

Thanks for reporting!

Do you mind trying Chrome? Deeplearn.JS is a new library and there are bound to be weird bugs like this I have no hope of debugging. I believe the most supported version is desktop Chrome.

Theoretically, for this algorithm, there is no size limit.

Practically, if you select something too high, your browser will die for lack of memory or you will grow old waiting for it to finish. My crappy 4GB RAM laptop can only handle around 300 x 500. Try setting the size to maximum, and if your computer can handle it, I'd be glad to guide you in using bigger images.

Okay, looked a bit into this and this piece of code is only run if an unsupported browser is detected, which is one of three cases, Mobile or Safari or a non-WebGL enabled browser.

I guess if you're not on mobile or Safari, then your browser doesn't support WebGL and you wouldn't be able to run it either. Sorry!

Aspect ratio shouldn't matter, the output should still be the same size as the Content image. Mind sharing the pic?

You took screenshots and it still broke? That's.. very odd, haha. I will look into it and see if it outputs out-of-range colors.

EDIT: My bad, there are indeed out of range colors! I will update here when fixed.

Hi, thanks for the report, but I'm not really a JS developer and have no idea how to work around this. Would you happen to know if there are any fixes I should look into?

This doesn't happen when I visit the site on Chrome on Linux. Do you mind sharing your system + browser?

Demo website: https://reiinakano.github.io/fast-style-transfer-deeplearnjs

Deeplearn.JS library doesn't work on mobile yet, sorry to folks on the go!

Demo link: https://reiinakano.github.io/fast-style-transfer-deeplearnjs...

Sure, but without seniority of the interviewer, would his/her reference carry as much weight?

Instead of banking on an interviewer's reputation, you would in essence be asking companies to trust your ability "to understand interviewer strength and what they are qualified to evaluate". Not so sure how well that'll work...

Having nobody you know at top tech companies to vouch for you is one of the biggest problems for people from other countries (esp. 3rd world countries) when applying. Can't even get through the resume screen.

This seems like a great way to provide a solution to this problem. Are you currently open to people who would need any form of visa sponsorship in the future? Assuming great performance, would they finally have a realistic shot at a job in a top tech company?

The two main differences are programming language (StackNet is Java) and the fact that Xcessiv is more a tool (heavily inspired by the flexibility of Jupyter notebook) and StackNet is a library.

Thanks for the feedback! Will do.

No. It is one curve per class.