|
Holy poo poo that's awesome!
|
#
?
Apr 17, 2015 15:41
|
|
|
|
| # ? Jun 11, 2024 10:35 |
|
|
Well, the software just batches videos so it's not worth posting a screenshot of, but a couple weeks ago I started working on video analysis to 'fake' a moving camera from a stationary one. results: http://i.imgur.com/6aeCShI.gif (caution, 7mb gif) The gif conversion made it sort of jumpy, it's actually nice and smooth. The next step is the analyze and fake arched movement so you get more of a fake human error\prediction curve. At this point I can do general height\distance output(for comparison, not accuracy).
|
|
#
?
Apr 18, 2015 17:52
|
|
|
Knyteguy posted:Quick update: Any particular reason you don't link to the Python code you referenced? Just curious as you refer to it as 'the Python software'. I'd definitely like to check the code out.
|
|
#
?
Apr 18, 2015 20:18
|
|
|
bobua posted:Well, the software just batches videos so it's not worth posting a screenshot of, but a couple weeks ago I started working on video analysis to 'fake' a moving camera from a stationary one. OpenCV? But yeah, it tracks him nicely, should work well if you fit it to a fitted curve maybe with some delayed "reaction"
|
|
#
?
Apr 18, 2015 21:21
|
|
|
Snapchat A Titty posted:OpenCV? Yeah. Getting it it done isn't all that hard but doing it fast is tough, especially since there tends to be so many moving objects in these shots. I basically have to scan the images, find the largest moving objects, go back through and remove outliers, go back through and get the start frame, direction of movement, highest frame, lowest frame after high frame, the plot the movement between those frames. Trying to just track the moving object produces some decent results most of the time but guarantees complete trash 10% of the time. Right now I still have an issue with the start frames Y axis every once in a while, but I'm thinking that's because in some cases the first X amount of frames don't contain the object I'm looking for.
|
|
#
?
Apr 19, 2015 01:26
|
|
|
Yeah the backtracking is a killer. I try to wrap it up to cache repeated functions, pre-sort & such, but even so it takes very little to suddently jump the complexity & used time.
|
|
#
?
Apr 19, 2015 01:32
|
|
|
MPEG video decoding using shaders. My first D3D11 app, lol. (Bitstream parse on the CPU, IDCT in a compute shader, rest in pixel shaders) http://i.imgur.com/E7mBIJj.jpg
|
|
#
?
Apr 19, 2015 03:12
|
|
|
Nice! This prodromal schizophrenic guy I met in a cafe challenged me to compute pi using OpenGL, but I can't use compute shaders because the laptop we agreed I could use doesn't support them :-(
|
|
#
?
Apr 19, 2015 03:16
|
|
|
Good for him I guess
|
|
#
?
Apr 19, 2015 03:22
|
|
|
TJChap2840 posted:Any particular reason you don't link to the Python code you referenced? Just curious as you refer to it as 'the Python software'. I'd definitely like to check the code out. Sure. There was no reason, no: https://github.com/scottrice/Ice I was incorrect saying that his implementation won't work in big picture (It looks like it will). What it won't do is work over streaming which was what I wanted. And here's my C# implementation (WIP): https://github.com/Noppadet/Emulation-Manager
|
|
#
?
Apr 19, 2015 03:55
|
|
|
My script language is now sufficiently expressive that I can implement quicksort with it:  Functions are also now first-class citizens: 
|
|
#
?
Apr 28, 2015 00:15
|
|
|
I wanted to make some triangle'd up images for my portfolio site but I didn't want to spend time working in Hexels or something, so I rolled my own little app.  20 hours of programming to save 2 hours of other work. ~An Engineering Solution~ It's up here if you want to play with it. I still have to clean up the code, optimize (its still pretty slow). When I have another stretch of free time I'll try to get it to make voronoi cells, too.
|
|
#
?
Apr 29, 2015 05:49
|
|
|
^^^ Thats really cool!
|
|
#
?
Apr 29, 2015 08:54
|
|
|
Thank you! I actually started and knocked out most of another project yesterday. I thought it was going to be way harder than it was, but they were pretty much all well-documented solved problems.  It's a little program that sits in your system tray and listens for PrintScreen presses (or double-clicks to its icon). It then lets you grab a partial screenshot with a selection rectangle, and immediately uploads it to imgur. In it's config you can set how you want it to mark the link up, so I can paste garbage screenshots in threads all day long with one keypress. It'll also ghetto-tweet (I'll probably actually work with the twitter API...later...) if you want it to. but look it's pretty fun!  once I get back from dinner I'll commit everything to git and post some download links e: up here if anyone wants to take a look. I'll probably finish up this weekend. Pentecoastal Elites fucked around with this message at 18:09 on May 1, 2015 |
|
#
?
May 1, 2015 12:40
|
|
|
Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: http://whatshouldicost.com Basically you connect your Github account and based on the languages you use and your length of experience, it'll give you a guesstimate of what your annual salary should be using data derived from the StackOverflow Dev Survey. Would be interesting to hear how accurate it is!
|
|
#
?
May 1, 2015 13:59
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: Interesting! Sadly it recommended I make 2.3x my current pay. Which is about right judging on what I've seen my friends get in the same areas. Luckily I finished my current job today..! 
|
|
#
?
May 1, 2015 14:02
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: Why do you need my email in the form? There's certificate errors when I try to go to the https site. Also,  Is that a bug? Or am I 
|
|
#
?
May 1, 2015 14:43
|
|
|
Drastic Actions posted:Why do you need my email in the form? There's certificate errors when I try to go to the https site. Also, I'm thinking about doing something a little bigger with it, making it an app to help with the recruitment process so the email is there so I can possibly alert people in the future. I'm working on it as a side project at my company which is trying to deal with eliminating recruiter spam so I should probably make it optional. quote:Is that a bug? Or am I Ha it was initially a problem with the NumberFormat component not showing enough precision but I left it in there because I thought it was pretty funny. Sticky Profits fucked around with this message at 14:56 on May 1, 2015 |
|
#
?
May 1, 2015 14:52
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this:  Apparently my experience in null is worth 4x more than I actually make though. 
|
|
#
?
May 1, 2015 15:07
|
|
|
SupSuper posted:Seems pretty accurate: You make 1/4 of a billion dollars?
|
|
#
?
May 1, 2015 15:59
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: It's VERY accurate, which is interesting considering I live in Chile. If I put the USA as my country, it shows 4x my salary. 
|
|
#
?
May 1, 2015 16:18
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: Seems quite accurate, though it thinks I should make less than I actually do. gmq posted:If I put the USA as my country, it shows 4x my salary. Same here  
|
|
#
?
May 1, 2015 16:27
|
|
|
netcat posted:Seems quite accurate, though it thinks I should make less than I actually do. Yeah, it's not a fan of my github account. Which is not too surprising considering that most of my work has been proprietary and most of my github stuff is test-free toy apps.
|
|
#
?
May 1, 2015 17:13
|
|
|
Interesting. Why can only pick New York, Silicon Valley, or Other as a US city?
|
|
#
?
May 1, 2015 18:37
|
|
|
  A data visualization tool for War in the Pacific; it reads in the data from the game logs and exports and generates this map. https://secure.baloogancampaign.com:8081/OperationGlacier/Map?side=Allies&date=411207
|
|
#
?
May 1, 2015 18:40
|
|
|
Factor Mystic posted:Interesting. Why can only pick New York, Silicon Valley, or Other as a US city? All(?) of the data it uses is from a StackOverflow survey; I assume there was just a lot more info available for those two cities than anywhere else.
|
|
#
?
May 1, 2015 21:34
|
|
|
kayakyakr posted:Yeah, it's not a fan of my github account. Which is not too surprising considering that most of my work has been proprietary and most of my github stuff is test-free toy apps. I wonder if startups would use it to figure out how much to pay foreign remote workers.
|
|
#
?
May 1, 2015 21:56
|
|
|
gmq posted:I wonder if startups would use it to figure out how much to pay foreign remote workers. The estimate I got was pretty much exactly the legal minimum wage for full time unskilled labor in Denmark. But then my github doesn't have a lot of public activity.
|
|
#
?
May 1, 2015 22:04
|
|
|
Darn, it gives me 60% of what I actually make. I need more stuff on GitHub.
|
|
#
?
May 2, 2015 01:24
|
|
Sticky Profits posted:Would be interesting to hear how accurate it is! I put in a bunch of "real" languages, including PHP and C#, as well as several hobby languages I use (Monkey, Blitzmax, some others) and it gave me $X. Out of curiosity, I took out all the "real" languages, and just left in the hobby languages, and $X actually went up by 10%. I am guessing it falls back on some sort of overall average if you enter nonsense data, like languages that probably did not show up in the survey.
|
|
|
#
?
May 2, 2015 10:04
|
|
|
 Working on a standalone podcast manager for ChromeOS (though it works on Windows/OSX/Linux too). Would be nice if different feed providers could stick to a standard format... AQB fucked around with this message at 23:03 on May 4, 2015 |
|
#
?
May 4, 2015 23:00
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: Location, location, location? I chose Other and it was nearly spot on but it's such a dramatic drop compared to Silicon Valley so I'm not sure if it is accurate. I'm in San Diego so shouldn't west coast Californian metropolitan area count for something or would Silicon Valley be the better option.
|
|
#
?
May 6, 2015 07:33
|
|
|
crazysim posted:Location, location, location? I chose Other and it was nearly spot on but it's such a dramatic drop compared to Silicon Valley so I'm not sure if it is accurate. I'm in San Diego so shouldn't west coast Californian metropolitan area count for something or would Silicon Valley be the better option. Also all the other Silicon ____'s out there have wages much more like Valley and less like bum gently caress Arkansas.
|
|
#
?
May 6, 2015 19:40
|
|
|
It said I was overpaid by about $18,000 but there also isn't a Seattle option on the ticker.  If it's any consolation, it would have been spot-on for my old job in Kentucky.
|
|
#
?
May 6, 2015 21:21
|
|
|
glompix posted:It said I was overpaid by about $18,000 but there also isn't a Seattle option on the ticker. I am sad, You can see the glass Ceiling from where i am - because i am squished against it.... My current salary is less than half what WSIC says i am worth... http://www.whatshouldicost.com/#/share/554af8b19bcb370f007ad86f
|
|
#
?
May 7, 2015 06:33
|
|
|
Sticky Profits posted:Not so much a screenshot as a working thing. Have been following an ML course and doing some React stuff lately and so built this: EDIT: I'm using git, of course. Because no Good Person would ever use Mercurial.  (I use Bitbucket because they were the only one offering free private repos for ages)
Shalinor fucked around with this message at 02:52 on May 8, 2015 |
|
#
?
May 7, 2015 16:04
|
|
|
Shalinor posted:This is cool, but I wish you'd expose your algorithm to manual entry. I'm assuming you're just pulling a handful of data from the Github, like language used and such? (I don't have a Github - Bitbucket supremacy! and all that) Wait... are you using Git on Bitbucket? Or Mercurial, like a Good Person? I jest; no intention to start a holy war here, I'm just thrilled when I find other people using Bitbucket by choice. Doubly so if they're using Mercurial too. To anyone about to continue this thread with a counterpost: no, I don't really care which DVCS and centralized (hah!) social hosting service you use...
|
|
#
?
May 8, 2015 00:33
|
|
|
I use SVN because it's worked for me so far in the past and I'm too old and lazy to bother changing. But now I don't know how much I should cost.
|
|
#
?
May 8, 2015 03:41
|
|
|
AntiPseudonym posted:I use SVN because it's worked for me so far in the past and I'm too old and lazy to bother changing. What I terrible mentality to have... The things that you can do with git make you so much more capable as a developer. You really should spend the 30 minutes or whatever that it takes to learn the basics.
|
|
#
?
May 8, 2015 03:49
|
|
|
|
| # ? Jun 11, 2024 10:35 |
|
|
I realized I hadn't written a neural network in a while, so I whipped up a multi-layer system. Here are three NNs, each with five layers containing 1, 10, 10, 10, and 1 nodes, trying to learn sin(x). The red one's activation function is the logit/sigmoid function. Green in softplus. Blue is tanh.
|
|
#
?
May 8, 2015 05:43
|
|
