BACK TO LONELINESS
BGANPUNKS V2 RARITY SANITY SH!THEYOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOI FORGOT TO MAKE THIS PAGE RESPONSIBLE, SO MOBILE USERS IF YOU ARE SEEING THINGS SUPER HUGE, SORRY. I AM USING A 2K SCREEN AND NOT A VERY PROFESSIONAL WEB DEV. GONNA FIX IT SOONFirst of all, words do not suffice to thank you all! I knew that people were gonna love BGANPUNKS V2, but I wouldn't expect the project to skyrocket at this pace. And the support messages your are sending are the sweetest...So since yesterday (the launch date), Imm'm getting a shittonnes of DMs what are the rarities of BASTARDS. I have some numbers I can share with you, on the other hand, about the visual appearance of BASTARDs, I might disappoint you, because in the process of generating them, there wasn't such a metric for rareness on my end. There were dozens of messy folders which had hundreds of iterations of more than 5 trained networks which had 4 different dataset variations etc. I played with renders and compressions to get different results for .GIFs, trained punks with both transparent and colored backgrounds, added my own custom backgrounds on some of them etc etc etc. But I hope that, from this text, you will understand, which BASTARDs might be rarer and more valuable.I AM SORRY THAT I HAVE NO STATISTICS IN MY HAND SAYING I HAVE X AMOUNT OF APES, ALIENS, X, Y, Z. GAN DOESN'T TELL ME SUCH STUFF WHILE GENERATING. IF I WANNA GET SOME DATA FROM GAN, I GET TENSORS WHICH ARE SHITTON OF NUMBERS WHICH ONLY MAKES SENSE TO THE ALGORITHM. I THINK YOUR EYES AND TASTE WILL BE THE DECISIVE FACTOR ON HOW VALUABLE THEY ARE. BECAUSE THE ALGORITHM DOES ITS BLACKMAGIC AND CREATE SOME WEIRD SHIT. THERE ARE SOMETIMES SUCH BIZARRE THINGS THAT I THINK IF WE HAD TO CHANCE WTF IS THIS TO THE GAN CODE, IT WOULD SAY IDKMostly upon generation of images, I told the script to pick random training points from the model to generate "X" BASTARDS. If you are seeing very pixelated, unrecognizable BASTARDS, it means that I generated those ones from a very early iteration of the training, which was just learning to distinguish the colors. I haven't put many of them as much as more recognizable ones because I thought that people would prefer to get BASTARDS that at least resemble a Punk. But after publishing the project I realized that many people adopting those very pixelated weird ones loved them and found them unique! So I can say in numbers they are rarer. (I can't say exact numbers because I don't have them in separate folders anymore).So let's start!
There are 11305 BASTARD GAN PUNK V2 in total.
847 of them are .GIF animations,
10458 of them are still .PNG images.
From .PNG images:* There are ones I trained with colored backgrounds which automatically generate bastards with a colored background.* Also other ones that were trained without a background (transparent), which were then layered with either a random RGB color, or a GPICPIX-GIF frame (that I will just explain a bit later)* As the result of some of my trained models, some traits weren't showing up for some reason, like hoodie, pink with hat, tassle hat etc. So I made an exclusive dataset including only those of with a couple more Cryptopunks traits. So there should be less of those. But I really can't don't remember how "less" they are because in order to count I need to look at all images one by one. I was doing lots of experimenting for weeks so don't take this as the absolute truth. This is how I remember it.The GLICPIX-GIF is another series of work of mine, which I will release some time in the future. They are basically .GIFs containing many frames which are pixelated glitchy 32x32 tiles. I thought that I can combine the .PNG layers of them randomly with some of the BASTARDS I generated from transparent background training.For around 1650-1770 still BASTARDS out of 10458, I layered my GLICPIX-.GIF layers with BASTARDS I generated with no background. GLIC-PIX is another series of mine which consist of lots of glitchy pixelated image tiles. So if you see a weird background instead of plain color, those are GLICPIX backgrounds. The number may not be very accurate but it definitely is pretty rare amongst all still images.In .GIF animations, I see there are around 160 ones which have very pixelated background animations (not a smooth single color transitioning). I think they are the most cool looking ones. And 25 of them I forgot to crop the edges lol, so the heads are in the middle of image and smaller. Does that make them rarer?Long story short, this is the info I could gather and remember for looks. If I come up with some stuff that I forgot, I will add them. Also don't forget this thing: I think all of them are unique in their way and GAN did very fascinating job. So once every one BASTARD is revealed you may find out that one of them has a super interesting unique look that I even hadn't noticed before. Maybe it was my mistake to not categorize looks (actually it would be much harder than just saying), but I also believe that it is not the thing in this project. All bastards are beautiful. All of them NEED LOVE. They will show you love as long as you show them love.
There are 11305 BASTARD GAN PUNK V2 in total.
847 of them are .GIF animations,
10458 of them are still .PNG images.
From .PNG images:* There are ones I trained with colored backgrounds which automatically generate bastards with a colored background.* Also other ones that were trained without a background (transparent), which were then layered with either a random RGB color, or a GPICPIX-GIF frame (that I will just explain a bit later)* As the result of some of my trained models, some traits weren't showing up for some reason, like hoodie, pink with hat, tassle hat etc. So I made an exclusive dataset including only those of with a couple more Cryptopunks traits. So there should be less of those. But I really can't don't remember how "less" they are because in order to count I need to look at all images one by one. I was doing lots of experimenting for weeks so don't take this as the absolute truth. This is how I remember it.The GLICPIX-GIF is another series of work of mine, which I will release some time in the future. They are basically .GIFs containing many frames which are pixelated glitchy 32x32 tiles. I thought that I can combine the .PNG layers of them randomly with some of the BASTARDS I generated from transparent background training.For around 1650-1770 still BASTARDS out of 10458, I layered my GLICPIX-.GIF layers with BASTARDS I generated with no background. GLIC-PIX is another series of mine which consist of lots of glitchy pixelated image tiles. So if you see a weird background instead of plain color, those are GLICPIX backgrounds. The number may not be very accurate but it definitely is pretty rare amongst all still images.In .GIF animations, I see there are around 160 ones which have very pixelated background animations (not a smooth single color transitioning). I think they are the most cool looking ones. And 25 of them I forgot to crop the edges lol, so the heads are in the middle of image and smaller. Does that make them rarer?Long story short, this is the info I could gather and remember for looks. If I come up with some stuff that I forgot, I will add them. Also don't forget this thing: I think all of them are unique in their way and GAN did very fascinating job. So once every one BASTARD is revealed you may find out that one of them has a super interesting unique look that I even hadn't noticed before. Maybe it was my mistake to not categorize looks (actually it would be much harder than just saying), but I also believe that it is not the thing in this project. All bastards are beautiful. All of them NEED LOVE. They will show you love as long as you show them love.
BASTARDNESS TRAIT RARITY TABLE
CLICK ON TYPES TO SEE THEM ON OPENSEA
UPDATE 09/03/2021. I realised that to fix something, I removed BIGBRAIN from the bastardness list instead. I noticed it after I got a report from a bastard adopter who had BIGBRAIN which mysteriously changed to 6HEAD. Sorry for these kind of amateur mistakes, having so much chaos and lack of sleep make me do these mistakes. I hope it will be the last.
DESCRIPTION RARITY TABLE
I wrote this short script to count how descriptions start. When I was generating the texts with GPT-2, I had an option to add "starting words". I choose different words or groups of words, but sometimes I just generated some stupid texts without giving them anything. In the table they are "RANDOM PHRASES". If I find out that I calculated the numbers wrong, BLAME THIS SHIT CODE
| DESCRIPTION STARTS WITH: | COUNT (SMALLER THE NUMBER, RARER IT IS) |
|---|---|
| RANDOM PHRASE | 1602 |
| DOOMED DEGEN | 1239 |
| BASTARD | 703 |
| CRYPTOPUNKS | 630 |
| BEING A NFT | 629 |
| BEING A BASTARD | 623 |
| REKT | 623 |
| BASTARD GAN PUNK | 622 |
| UGLY STINKY WHALES | 606 |
| CRYPTO | 604 |
| BOOMER | 603 |
| PUNK LIFE | 583 |
| FUCKING | 567 |
| SHIT | 566 |
| NON FUNGIBLE TOKENS | 561 |
| DEGEN | 544 |