CLIP-GEN: Language-Free Training of a Text-to-Image Generator with CLIP

Overview

CLIP-GEN

[简体中文][English]

本项目在萤火二号集群上用 PyTorch 实现了论文 《CLIP-GEN: Language-Free Training of a Text-to-Image Generator with CLIP》。

clip-gen

CLIP-GEN 是一个 Language-Free 的文本生成图像的方法,它不依赖图文训练样本,通过预训练 CLIP 模型的强大表征能力,只需要图片数据就可以训练出一个文本生成图像的模型。该方法的基本原理是:CLIP-GEN 首先会训练一个 VQ-GAN,把图片映射到离散空间;然后再训练一个 GPT 模型,把 CLIP embedding 映射到 VQ-GAN 的离散空间;由于在 CLIP 中,文本和图像共享一个特征空间,在 inference 的时候我们就可以通过同样的方法把文本映射到 VQ-GAN 的离散空间,然后 decode 为 RGB 图像。

Requirements

  • hfai (to be released soon)
  • torch>=1.8

Training

支持的数据集:coco, imagenet, googlecc

  1. 下载 CLIP 预训练模型

    下载 CLIP 后放至 pretrained/clip_vit_b32.pt,该预训练模型来自 OpenAI.

  2. 在 COCO 上训练 VQGAN

    提交任务至萤火集群:

    hfai python train_vqgan.py --ds coco -- -n 1 -p 30

    本地运行:

    python train_vqgan.py --ds coco
  3. 在 COCO 上训练 Conditional GPT

    提交任务至萤火集群:

    hfai python train_gpt.py --ds coco --vqgan_ckpt /path/to/vqgan/ckpt -- -n 4 -p 30

    本地运行:

    python train_gpt.py --ds coco --vqgan_ckpt /path/to/vqgan/ckpt

Demo

下载在 COCO 上训练好的 VQGANGPT 模型,分别放到 pretrained/vqgan_coco.ptpretrained/gpt_coco.pt;然后运行:

python demo.py --text "A city bus driving on the city street" --out "bus.jpg"

NOTE: demo 的运行不依赖 hfai,用户可以在装有 PyTorch 的环境下直接使用

Samples

下面是一些文本生成图像的样本:

tower bus living train skiing

References

Citation

@article{wang2022clip,
  title={CLIP-GEN: Language-Free Training of a Text-to-Image Generator with CLIP},
  author={Wang, Zihao and Liu, Wei and He, Qian and Wu, Xinglong and Yi, Zili},
  journal={arXiv preprint arXiv:2203.00386},
  year={2022}
}

TODO

  • 预训练模型
  • FFRecord 数据
You might also like...
Zero-Shot Text-to-Image Generation VQGAN+CLIP Dockerized
Zero-Shot Text-to-Image Generation VQGAN+CLIP Dockerized

VQGAN-CLIP-Docker About Zero-Shot Text-to-Image Generation VQGAN+CLIP Dockerized This is a stripped and minimal dependency repository for running loca

A Jupyter notebook to play with NVIDIA's StyleGAN3 and OpenAI's CLIP for a text-based guided image generation.

A Jupyter notebook to play with NVIDIA's StyleGAN3 and OpenAI's CLIP for a text-based guided image generation.

Text2Art is an AI art generator powered with VQGAN + CLIP and CLIPDrawer models
Text2Art is an AI art generator powered with VQGAN + CLIP and CLIPDrawer models

Text2Art is an AI art generator powered with VQGAN + CLIP and CLIPDrawer models. You can easily generate all kind of art from drawing, painting, sketch, or even a specific artist style just using a text input. You can also specify the dimensions of the image. The process can take 3-20 mins and the results will be emailed to you.

A 1.3B text-to-image generation model trained on 14 million image-text pairs
A 1.3B text-to-image generation model trained on 14 million image-text pairs

minDALL-E on Conceptual Captions minDALL-E, named after minGPT, is a 1.3B text-to-image generation model trained on 14 million image-text pairs for no

A PyTorch Lightning solution to training OpenAI's CLIP from scratch.
A PyTorch Lightning solution to training OpenAI's CLIP from scratch.

train-CLIP 📎 A PyTorch Lightning solution to training CLIP from scratch. Goal ⚽ Our aim is to create an easy to use Lightning implementation of OpenA

[IJCAI-2021] A benchmark of data-free knowledge distillation from paper
[IJCAI-2021] A benchmark of data-free knowledge distillation from paper "Contrastive Model Inversion for Data-Free Knowledge Distillation"

DataFree A benchmark of data-free knowledge distillation from paper "Contrastive Model Inversion for Data-Free Knowledge Distillation" Authors: Gongfa

Free-duolingo-plus - Duolingo account creator that uses your invite code to get you free duolingo plus
Free-duolingo-plus - Duolingo account creator that uses your invite code to get you free duolingo plus

free-duolingo-plus duolingo account creator that uses your invite code to get yo

Official implementation of SynthTIGER (Synthetic Text Image GEneratoR) ICDAR 2021
Official implementation of SynthTIGER (Synthetic Text Image GEneratoR) ICDAR 2021

🐯 SynthTIGER: Synthetic Text Image GEneratoR Official implementation of SynthTIGER | Paper | Datasets Moonbin Yim1, Yoonsik Kim1, Han-cheol Cho1, Sun

The source code for Generating Training Data with Language Models: Towards Zero-Shot Language Understanding.
The source code for Generating Training Data with Language Models: Towards Zero-Shot Language Understanding.

SuperGen The source code for Generating Training Data with Language Models: Towards Zero-Shot Language Understanding. Requirements Before running, you

Comments
  • "nn.TransformerEncoderLayer" is adopted to construct the "conditonal transformer" in your paper.

    Thanks for your great work.

    I noticed that you utilize "nn.TransformerEncoderLayer" when constructing "conditional transformer". Since it is used to predict the next token index, I am wondering whether the decoder of transformer is more appropriate for the construction of your conditional transformer? or what's the reason that you don't adopt "nn.TransformerdecoderLayer" ?

    Because of the structure of "nn.TransformerEncoderLayer" is simpler or more concise than that of "nn.TransformerDEcoderLayer" ?

    opened by fido20160817 0
  • Add Web Demo & Docker environment

    Add Web Demo & Docker environment

    This pull request makes it possible to run your model inside a Docker environment, which makes it easier for other people to run it. We're using an open source tool called Cog to make this process easier.

    This also means we can make a web page where other people can try out your model, view it here: https://replicate.com/hfailab/clip-gen. You can find the docker file under the tab ‘run model with docker’.

    We have added some examples to the page, but do claim the page so you can own the page, customise the Example gallery as you like, push any future update to the web demo, and we'll feature it on our website and tweet about it too. You can find the 'Claim this model' button on the top of the page. Any member of the HFAiLab organization on GitHub can claim the model ~ When the page is claimed, it will be automatically linked to the arXiv website as well (under “Demos”).

    In case you're wondering who I am, I'm from Replicate, where we're trying to make machine learning reproducible. We got frustrated that we couldn't run all the really interesting ML work being done. So, we're going round implementing models we like. 😊

    opened by chenxwh 0
Owner
null
CLIP (Contrastive Language–Image Pre-training) trained on Indonesian data

CLIP-Indonesian CLIP (Radford et al., 2021) is a multimodal model that can connect images and text by training a vision encoder and a text encoder joi

Galuh 17 Mar 10, 2022
CLIP: Connecting Text and Image (Learning Transferable Visual Models From Natural Language Supervision)

CLIP (Contrastive Language–Image Pre-training) Experiments (Evaluation) Model Dataset Acc (%) ViT-B/32 (Paper) CIFAR100 65.1 ViT-B/32 (Our) CIFAR100 6

Myeongjun Kim 52 Jan 7, 2023
PyTorch implementation of 'Gen-LaneNet: a generalized and scalable approach for 3D lane detection'

(pytorch) Gen-LaneNet: a generalized and scalable approach for 3D lane detection Introduction This is a pytorch implementation of Gen-LaneNet, which p

Yuliang Guo 233 Jan 6, 2023
Code for our CVPR 2022 Paper "GEN-VLKT: Simplify Association and Enhance Interaction Understanding for HOI Detection"

GEN-VLKT Code for our CVPR 2022 paper "GEN-VLKT: Simplify Association and Enhance Interaction Understanding for HOI Detection". Contributed by Yue Lia

Yue Liao 47 Dec 4, 2022
TAP: Text-Aware Pre-training for Text-VQA and Text-Caption, CVPR 2021 (Oral)

TAP: Text-Aware Pre-training TAP: Text-Aware Pre-training for Text-VQA and Text-Caption by Zhengyuan Yang, Yijuan Lu, Jianfeng Wang, Xi Yin, Dinei Flo

Microsoft 61 Nov 14, 2022
PyTorch code for BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation

BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation

Salesforce 1.3k Dec 31, 2022
Simple command line tool for text to image generation using OpenAI's CLIP and Siren (Implicit neural representation network)

Deep Daze mist over green hills shattered plates on the grass cosmic love and attention a time traveler in the crowd life during the plague meditative

Phil Wang 4.4k Jan 3, 2023
CLIP+FFT text-to-image

Aphantasia This is a text-to-image tool, part of the artwork of the same name. Based on CLIP model, with FFT parameterizer from Lucent library as a ge

vadim epstein 690 Jan 2, 2023
A simple command line tool for text to image generation, using OpenAI's CLIP and a BigGAN.

Ryan Murdock has done it again, combining OpenAI's CLIP and the generator from a BigGAN! This repository wraps up his work so it is easily accessible to anyone who owns a GPU.

Phil Wang 2.3k Jan 9, 2023
Source code for models described in the paper "AudioCLIP: Extending CLIP to Image, Text and Audio" (https://arxiv.org/abs/2106.13043)

AudioCLIP Extending CLIP to Image, Text and Audio This repository contains implementation of the models described in the paper arXiv:2106.13043. This

null 458 Jan 2, 2023