//=time() ?>
ControlNetのSegmentationで、ポジティブプロンプトを一切入力せずに自作のセグメンテーション画像だけで構図を作るテスト
使われるセグメンテーションの色とオブジェクトの関係さえわかっちゃえば一般的な構図は自由に作れそう
NPはEasyNegative, (worst quality, low quality:1.4)
#AIイラスト
Control Stable Diffusion with Segmentation Maps #ControlNet
"artwork by {artistname}, stunning city landscape, street view, detailed"
Eyvind Earle https://t.co/waxFx7g98h
John Berkey https://t.co/t6gvvf7eHp
Alphonso Mucha https://t.co/rxVmU5QUQl
ControlNet is pretty wild 🔥 First image generated using #midjourney and then converted using #controlnet HED, Canny and Scribble. While the other two are cool for converting style, Scribble and Segmentation (not pictured) seem to be great ways to transform but keeping structure.
アニメ画像の背景消すツール作った🛠
https://t.co/1rBecQ7Neo
(pip一発で入れられるようにしたSkyTNT/anime-segmentationのラッパー)
anime-segmentationとbizarre-pose-estimatorを併用すると多少骨格の推定精度が上がることを確認
かなり難しい画像でやってるけど、こんな服でこれだけいけるなら、十分だろう
昨日と今朝で、切り抜きツールanime-segmentation と i2iで、うちのこが堕ちる前の差分作ってました
クッソかわいい……女神や……
※堕ちシーンはトップツイート参照
#AIイラスト
Unofficial Implementation of Paint-with-words, method from eDiffi that let you generate image from text-labeled segmentation map
github: https://t.co/4VbbU9Mdyv
First look on my upcoming project on Musteloidea endocast shape!🧠🦦 And still a lot of segmentation to do for the future 😅
The goat lady is so funny, i only intended her to break the 4ty wall once with her segmentation power
But here i just accidentally made her too big lol
Why does my "life-ambience" suddenly shift without warning and inflict jagged segmentation between memories and episodes of my life? Why can't it just be a smooth and seamless transition that feels compounding and organic? Is this normal?
新一お誕生日おめでとう🥳💛
【web再録】segmentation【快新】【K新】
2019年の作品、当時手に取ってくださった方々
ありがとうございました!
(なんとなく恥ずかしくなったら下げます)
全文はコチラ→https://t.co/Vj4KFt0kP5
※快K新の表現があるので注意
①↓次頁の見開きとか当時がんばったw
smooches him. idk im still having a design crisis over deja vu's whole thing...I'm attached to his noodley arms but this style also makes me :eyes:
the segmentation of his boots is inevitable since he needs to bend his feet after all..still not super set on anything tho
it's also interesting (..at least for me, idk for others) to see how simple characters designs choice for that anime era too (even the wackier ones like five star stories, where they emphasize shapes over random swirls/segmentation)
Deep learning-based segmentation of the thorax in mouse micro-CT scans https://t.co/qO0GrxZbjH
Another new collection in progress, Cyberpunk Apocalypse, a series mixing #vqganclip landscapes with StyleGAN2+CLIP avatars. Here they are separately, working on image segmentation to combine them!
#NFTcommunity #AIArtwork #generativeart #WorkinProgress