If you would like to comment you will need to be logged in. Register now. It only takes a moment and it's free.

b236 (Level 6 Time Traveller) @ 5/1/2021 00:03


StoneStephenT (Level 8 Ruby) @ 4/30/2021 21:53



Hapiel (Level 11 Pixellite) @ 4/23/2021 02:22

Mythical creatures, you picked a good one! Here is more info about the introduction of the new themes :)

Goodlyay (Level 6 Centaur) @ 4/23/2021 00:15

Yoo, when did PJ get more rank themes? They look great. Can't figure out how to change mine anymore, though...
Got it.

gawrone (Level 8 Nebula) @ 4/16/2021 03:44

Hapiel. Late but still 🎂🎂🎂

Hapiel (Level 11 Pixellite) @ 4/16/2021 00:59

Thank you!!

cmmrc (Level 4 Button Man) @ 4/15/2021 16:26

Happy birthday Hapiel!

Hapiel (Level 11 Pixellite) @ 3/28/2021 23:38

But don't you influence it already with your choice to pick a 3x3 tile? And I'm not suggesting we handpick the data, just to have a look at those inbetween artworks to see what the result is, because now you're blindly assuming that your set is selecting on dithering, even though it might pick up a lot of noise!

Motrox (Level 3 Private Eye) @ 3/28/2021 21:54


I see what you mean. Thing about hand selecting a dataset is that I don't want to introduce my own biases.

It would be nice if there was a dataset that is split between "dithering" and "non-dithering", to run some experiments on. Somehow need to crowdsource I guess. I do know that the PJ dataset has tags, so I can filter for dithered images, I hope. 

I will definitely message you more about this. I'm excited for a rigourous analysis now. I'll get some more graphs for you on the other patches if you're interested!

Motrox (Level 3 Private Eye) @ 3/28/2021 21:50

That's an interesting way to look at it. There is a lot that has been said on noise estimation, so that is something I will look into. 

Hapiel (Level 11 Pixellite) @ 3/28/2021 05:47

If you want to know how often it fails:
Try to make a list of artwork that has a 2x3 or 3x2 patch but not a 3x3 patch. That should be easy to do, and then we can look over these together to see what we think of the works, whether we consider them to have dithering or not.

Similarly, I'd be interested in a list of works that have 3x3 but not 4x4 patches, so we can check if a lot of them have random x spots or if they have dithering but just minimal.

Also, still hoping to see the 4x4 and 5x5 graphs to compare, to see if the curve is the same. My instinct is that larger patches are really something of the past, whereas small patches are more common in present day too.

Hapiel (Level 11 Pixellite) @ 3/28/2021 05:43

It's buggy for me too, I might try later with fixeljoint turned off perhaps thats it! But opening the fav link in a new tab has solved it for me, that way I can still fav.

b236 (Level 6 Time Traveller) @ 3/28/2021 00:26

@Motrox: i have a "signature" which contains exactly that patch. :D letter "x" is another case.
anyway, considering dithering is technically a noise, shouldn't be the best option to estimate amount of noise in each image? you wouldn't need machine learning for that.

btw. i was surprised how many images actually contain dithering even in this simplest form.

Motrox (Level 3 Private Eye) @ 3/27/2021 11:23

The way I'm searching right now is not be fit to find the coverage of dithering. 

I was thinking about this a little more, and coverage seems like a more telling attribute to analyze, rather than classifying the image as containing simple dithering or not. 

Motrox (Level 3 Private Eye) @ 3/27/2021 11:02

The program looks for a 3x3 patch of dithering within the image, like Hapiel said. Dithering blocks come in all sorts of shapes, so I  decided to target what is probable to be common between all of them; Also, I reduced the scope to "simple dithering" rather than stylized because I was not sure how to deal with all the variations. Some places it can fail:

1.) One place it will fail to detect is if the image only contains slight dither less than the 3x3. My assumption is this is quite rare. What do you think of that? 

2.) One place it will misclassify as dithering is If the artist added the 3x3 patch, but did not intend for it to be interpreted as dithering. My thinking is that this is also unlikely. Can you find an example of this?

I will look into how to evaluate the goodness of a binary classifier to get some numbers for this method. Machine learning is something I don't know too well, but that is also another possibility. 

I'm definitely interested in looking into how techniques have changed further. An analysis of all fundamental methods in pixelart would be interesting.

CELS (Level 7 Cuttlefish) @ 3/27/2021 07:47

Yeah, the evolution of the pixel art style in the last 10 years would be fascinating to look at. Although if one looks at popular indie pixel art games from the last 2-3 years, it's definitely hard to identify universal trands. I guess one would have to separate the art scene from the games, for a start. 

Gecimen (Level 11 Sphinx) @ 3/27/2021 07:41

Academic studies on pixel art is next to non existent so would be an awesome place to start.

CELS (Level 7 Cuttlefish) @ 3/27/2021 07:27

Thanks for the detailed explanation! Sounds like a job for someone writing a thesis on pixel art. :) 

cmmrc (Level 4 Button Man) @ 3/27/2021 07:11

I haven't been able to add anything to my favorites for a while, might that be a fixeljoint extension issue maybe?

Hapiel (Level 11 Pixellite) @ 3/27/2021 06:52
knowing how Motrox analised the data, I think it would be very hard to check the total dithered area. He searches for a small patch of 3x3 checkerboard, but that method would not work very well on the example you linked. With many areas in that image it would be really hard to decide if it is dithering or not, especially if you were to view it in isolation...
Like in this snippet, I marked some areas that the algorithm would detect with red:

But the areas marked blue wouldn't be detected, as it isn't a perfect 3x3 checkerboard. But in this artwork, if you were to zoom in on any of the areas marked orange, would you be able to say that it is dithering or not? Where does the dithering begin, and where does it end? In the context of the piece, seeing the other dithered areas connected to it makes it easier to say that the piece has 'a lot' of dithering, but that is probably much more difficult to train a computer to recognise.

You perhaps could try and train a machine learning algorithm, but then you need training data meaning that some humans should rate different artworks by their amount of ditheredness.. And I wouldn't want to do that, because I wouldn't know how to count, haha

CELS (Level 7 Cuttlefish) @ 3/27/2021 04:55

That's interesting! Presumably, you've made some kind of algorithm that does the calculations for you. Would it be a complicated task to expand this to investigate the percentage of artworks with more than 5% of the total area devoted to dithering? For example, vierbit's latest submarine in a bottle uses a tiny bit of dithering. But there's a lot more dithering in his very first piece from 12 years ago. And roughly the same number of colors.

Motrox (Level 3 Private Eye) @ 3/26/2021 15:36

My analysis of simple dithering technique (checkerboard pattern) on Pixeljoint, given as percentages of artworks that use this technique for each year:


Wolfenoctis (Level 8 Cannibal) @ 3/19/2021 04:17

Thanks DioShiba

DioShiba (Level 4 Shoggoth (The Mountains of Madness)) @ 3/18/2021 15:21

Happy Birthday Wolfenoctis!

Wolfenoctis (Level 8 Cannibal) @ 3/17/2021 22:01

Thanks Theoden

Theoden (Level 8 Marshal) @ 3/17/2021 13:47

Happy birthday Wolfenoctis!

Wolfenoctis (Level 8 Cannibal) @ 3/17/2021 07:42

Thanks Hapiel

Hapiel (Level 11 Pixellite) @ 3/17/2021 05:39

Happy birthday!!

Wolfenoctis (Level 8 Cannibal) @ 3/17/2021 04:43

Thank you Irenaart

Irenaart (Level 8 Marshal) @ 3/17/2021 03:48

@Wolfenoctis; Happy birthday, man! 1st place in monthly top 10 is some kind of birthday present, ha... Well deserved!

wakimojo (Level 3 Dodo) @ 3/16/2021 15:03

i'd keep them all, and add more. there are already 8 sets, why not 20 or 30. in this case i suppose i'd rather have 1-3 or 100 sets. each artist is unique with their individual proclivities, so i don't see a problem with there being too many choices. it's just for fun, and many of the rankings don't indicate level very coherently anyway.

Eggy (Level 5 Detective) @ 3/16/2021 05:04

I really hope Law and Order doesn't get deleted, lol. When I first joined, that was the only rank theme I liked out of all the ones available and it's grown on me since. I like the idea of being a pretend cop 

b236 (Level 6 Time Traveller) @ 3/15/2021 16:47

in case of inevitable: farewell, my army ranks, you were fake and that's what i liked about you.

gawrone (Level 8 Nebula) @ 3/15/2021 12:25

Everyone have the right to be wrong

StevenM (Level 6 GBC) @ 3/15/2021 11:41

Hard disagree

gawrone (Level 8 Nebula) @ 3/15/2021 10:53

Nah, they're good. And it's the coolest and the best looking set among them.

StevenM (Level 6 GBC) @ 3/15/2021 10:09

The only good Psycho icon is the Jason mask

skeddles (Level 11 Sphinx) @ 3/15/2021 08:29

if you delete psycho im deleting my account

Hapiel (Level 11 Pixellite) @ 3/15/2021 06:34

Currently in the discord we're working on some new ranking icon sets, and were discussing how these could be implemented in PJ soon. If you still weren't sure whether to join Discord or not, perhaps this can be your reason (link in the menu bar under "Community")

CELS (Level 7 Cuttlefish) @ 3/12/2021 01:57

Yeah, it looks brilliant! I just hope it has more depth than some of the old beat em ups. They were usually about mastering a few simple moves. And it's made by the people who did Streets of Rage 4, which is a fairly simplistic (but pretty) game, so I'm a bit worried that I'm going to have to button mash for 3 hours just to see all the pretty pixel art. 

StoneStephenT (Level 8 Ruby) @ 3/10/2021 10:04

That new TMNT game that got revealed today looks awesome, graphics-wise. Feels like a natural evolution of the old Konami games. The Turtles even have their own individual animations!

Hapiel (Level 11 Pixellite) @ 3/8/2021 06:30

glad that I could help :p

gawrone (Level 8 Nebula) @ 3/8/2021 05:58

Thanks Hapiel! Now I feel silly that I used PJ Image Specs for colourcount only, and never noticed it has exactly what I described :| and also becasue Fixeljoint is new to me. I will never look back on how my life was difficult before.

Hapiel (Level 11 Pixellite) @ 3/8/2021 05:40
The best solution is to use the Fixeljoint browser extension. It adds a link on every page to the PJ image specs viewer, which allows for watching animations frame by frame!
I wouldn't know what to do without it!

gawrone (Level 8 Nebula) @ 3/8/2021 05:11

I have an idea for a PJ feature, but I'm not very clever in coding, so please don't diss me. The greatest thing about PJ for me is the opportunity to learn from amazing artists, by peeking on how they solve particular problems. And for pixel art there could not be anything better as clicking + to enlarge the details.

Could we have something similar for animation? The ability to stop the gif to examine one frame or to move between individual frames? I know, that I could just dowload the giff and open it, but the same thing could be said about enlarging and yet we have it build in and it is amazing.

SeDiceBisonte (Level 6 Blue Crab) @ 3/8/2021 03:01

I wanted to give the demo a try but was worried I'd missed it. Suspected there may be a workaround since it still shows up in Steam's search results, and it turns out you can make it install with this link: steam://install/1519390

Hopefully that helps some people who want to try before buying. I need to force myself to do more pixelling before actually playing it (but I can confirm the demo installed).

DimDim (Level 5 MUTO) @ 3/6/2021 10:05

I think I downloaded it a while ago when they first showed it, but never got to play it, which happens often. If you say it's not really a card game I'll take a look, maybe I can learn something from those pixels.

skamocore (Level 11 Dolphin) @ 3/6/2021 00:35

As someone who's also generally turned off by card-based games, this really isn't a card game. Think of it as your inventory being being represented by rectangular images. But, then again, what really makes something a card game? whoooo knooows.

It's a shame they removed the demo from Steam.

DimDim (Level 5 MUTO) @ 3/5/2021 16:27

That gritty art is mesmerizing

But is a card based game so I sleep

skamocore (Level 11 Dolphin) @ 3/5/2021 01:16

I'm loving the game so far. Also, art by Deceiver.

of 20

This Week's Pixel Art

Luminescent Reaper icon/pixelartCrab Rider icon/pixelartSix Fanarts icon/pixelartOrc Dagger of Skyrim (Well, an attempt from her). icon/pixelart
Play party games with QuizBash app
Play party games with QuizBash app

Random Pixel Art

Super Maid-rio café! icon/pixelart
Vintage Wooden Radio icon/pixelart


Want to give some dough back to all those amazing pixel artists? Donations provide prize money for contests, help cover hosting costs and support new initiatives.