# 文本轉圖片 ###### tags: `文本轉圖片` [Jamboard](https://jamboard.google.com/d/1X1fs-qOCcFpBhqyrq4JL8hnYXJI4_xwINu3drUnrTF8/edit?usp=sharing) ## Papers [**StoryGAN: A Sequential Conditional GAN for Story Visualization**](https://openaccess.thecvf.com/content_CVPR_2019/html/Li_StoryGAN_A_Sequential_Conditional_GAN_for_Story_Visualization_CVPR_2019_paper.html) 被引用: 63 [Improved-StoryGAN for sequential images visualization](https://www.sciencedirect.com/science/article/abs/pii/S1047320320301826) 被引用: 1 [StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks](https://arxiv.org/abs/1612.03242) 被引用:1871 > 原有的算法無法對文本轉圖片提供足夠的細節,StackGAN將其分為兩個子問題,階段1先根據 > 給定的文本描繪出低畫質的輪廓及顏色,第2階段再以文本與第1階段的圖案當作輸入,生成具 > 細節的圖像(使用了 novel Conditioning Augmentation technique) [StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks](https://arxiv.org/abs/1710.10916) 被引用:590 > 由多個生成器與樹狀的鑑別器組成,從樹的不同分支生成對應於相同場景的不同解析度的圖像 [Generative Adversarial Text to Image Synthesis](http://proceedings.mlr.press/v48/reed16.html) 被引用: 2236, Submitted on 17 May 2016 > 就是最一開始提出效果還不錯的text to image的那篇論文 [Lightweight dynamic conditional GAN with pyramid attention for text-to-image synthesis](https://www.sciencedirect.com/science/article/pii/S0031320320301874) 被引用: 10, Submitted on 2021 > 為了生成高解析度的圖像,神經網路參數和復雜性急劇增加,造成訓練網路過程不穩定和訓練成本高的問題。為了解決這個問題本文提出一個LD-CGAN(一個複雜的結構)。與當前最先進的 HDGAN 相比,LD-CGAN 分別顯著減少了 86.8% 和 94.9% 的參數數量和計算時間。 [AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks](https://arxiv.org/abs/1711.10485) ## 流程 文本 -> 摘要 -> StoryGAN ## Books [暮光之城](https://www.youduzw.com/tags/13/) https://online.pubhtml5.com/pien/hbbk/#p=4
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up