File size: 7,571 Bytes
6b10b56
 
f373c9c
 
6b10b56
f373c9c
8d6a880
6b10b56
f373c9c
 
 
 
 
 
 
 
 
 
 
 
 
 
da59868
f373c9c
 
 
 
3aeff9f
f373c9c
 
3aeff9f
f373c9c
 
3aeff9f
 
 
f373c9c
 
 
456ce74
f373c9c
456ce74
3aeff9f
f373c9c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3aeff9f
 
 
f373c9c
 
 
4712e8a
f373c9c
 
 
 
 
456ce74
f373c9c
 
 
 
3aeff9f
 
f373c9c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3aeff9f
 
 
f373c9c
 
 
 
 
 
 
 
 
 
3aeff9f
f373c9c
3aeff9f
 
f373c9c
3aeff9f
 
 
 
f373c9c
4712e8a
f373c9c
 
3aeff9f
f373c9c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
---
license: apache-2.0
prior:
- kandinsky-community/kandinsky-2-1-prior
tags:
- text-to-image
- kandinsky
---

# Kandinsky 2.1

Kandinsky 2.1 inherits best practices from Dall-E 2 and Latent diffusion while introducing some new ideas.

It uses the CLIP model as a text and image encoder,  and diffusion image prior (mapping) between latent spaces of CLIP modalities. This approach increases the visual performance of the model and unveils new horizons in blending images and text-guided image manipulation.

The Kandinsky model is created by [Arseniy Shakhmatov](https://github.com/cene555), [Anton Razzhigaev](https://github.com/razzant), [Aleksandr Nikolich](https://github.com/AlexWortega), [Igor Pavlov](https://github.com/boomb0om), [Andrey Kuznetsov](https://github.com/kuznetsoffandrey) and [Denis Dimitrov](https://github.com/denndimitrov)

## Usage

Kandinsky 2.1 is available in diffusers!

```python
pip install diffusers transformers accelerate
```
### Text to image

```python
from diffusers import DiffusionPipeline
import torch

pipe_prior = DiffusionPipeline.from_pretrained("kandinsky-community/kandinsky-2-1-prior", torch_dtype=torch.float16)
pipe_prior.to("cuda")

t2i_pipe = DiffusionPipeline.from_pretrained("kandinsky-community/kandinsky-2-1", torch_dtype=torch.float16)
t2i_pipe.to("cuda")

prompt = "A alien cheeseburger creature eating itself, claymation, cinematic, moody lighting"
negative_prompt = "low quality, bad quality"

image_embeds, negative_image_embeds = pipe_prior(prompt, negative_prompt, guidance_scale=1.0).to_tuple()

image = t2i_pipe(prompt, negative_prompt=negative_prompt, image_embeds=image_embeds, negative_image_embeds=negative_image_embeds, height=768, width=768).images[0]
image.save("cheeseburger_monster.png")
```

![img](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/kandinsky-docs/cheeseburger.png)


### Text Guided Image-to-Image Generation

```python
from diffusers import KandinskyImg2ImgPipeline, KandinskyPriorPipeline
import torch

from PIL import Image
import requests
from io import BytesIO

url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
response = requests.get(url)
original_image = Image.open(BytesIO(response.content)).convert("RGB")
original_image = original_image.resize((768, 512))

# create prior
pipe_prior = KandinskyPriorPipeline.from_pretrained(
    "kandinsky-community/kandinsky-2-1-prior", torch_dtype=torch.float16
)
pipe_prior.to("cuda")

# create img2img pipeline
pipe = KandinskyImg2ImgPipeline.from_pretrained("kandinsky-community/kandinsky-2-1", torch_dtype=torch.float16)
pipe.to("cuda")

prompt = "A fantasy landscape, Cinematic lighting"
negative_prompt = "low quality, bad quality"

image_embeds, negative_image_embeds = pipe_prior(prompt, negative_prompt).to_tuple()

out = pipe(
    prompt,
    image=original_image,
    image_embeds=image_embeds,
    negative_image_embeds=negative_image_embeds,
    height=768,
    width=768,
    strength=0.3,
)

out.images[0].save("fantasy_land.png")
```

![img](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/kandinsky-docs/img2img_fantasyland.png)


### Interpolate 

```python
from diffusers import KandinskyPriorPipeline, KandinskyPipeline
from diffusers.utils import load_image
import PIL

import torch

pipe_prior = KandinskyPriorPipeline.from_pretrained(
    "kandinsky-community/kandinsky-2-1-prior", torch_dtype=torch.float16
)
pipe_prior.to("cuda")

img1 = load_image(
    "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/kandinsky/cat.png"
)

img2 = load_image(
    "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main" "/kandinsky/starry_night.jpeg"
)

# add all the conditions we want to interpolate, can be either text or image
images_texts = ["a cat", img1, img2]

# specify the weights for each condition in images_texts
weights = [0.3, 0.3, 0.4]

# We can leave the prompt empty
prompt = ""
prior_out = pipe_prior.interpolate(images_texts, weights)

pipe = KandinskyPipeline.from_pretrained("kandinsky-community/kandinsky-2-1", torch_dtype=torch.float16)
pipe.to("cuda")

image = pipe(prompt, **prior_out, height=768, width=768).images[0]

image.save("starry_cat.png")
```
![img](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/kandinsky-docs/starry_cat.png)


## Model Architecture

### Overview
Kandinsky 2.1 is a text-conditional diffusion model based on unCLIP and latent diffusion, composed of a transformer-based image prior model, a unet diffusion model, and a decoder.   

The model architectures are illustrated in the figure below - the chart on the left describes the process to train the image prior model, the figure in the center is the text-to-image generation process, and the figure on the right is image interpolation. 

<p float="left">
  <img src="https://raw.githubusercontent.com/ai-forever/Kandinsky-2/main/content/kandinsky21.png"/>
</p>

Specifically, the image prior model was trained on CLIP text and image embeddings generated with a pre-trained [mCLIP model](https://huggingface.co/M-CLIP/XLM-Roberta-Large-Vit-L-14). The trained image prior model is then used to generate mCLIP image embeddings for input text prompts. Both the input text prompts and its mCLIP image embeddings are used in the diffusion process. A [MoVQGAN](https://openreview.net/forum?id=Qb-AoSw4Jnm) model acts as the final block of the model, which decodes the latent representation into an actual image.


### Details
The image prior training of the model was performed on the [LAION Improved Aesthetics dataset](https://huggingface.co/datasets/bhargavsdesai/laion_improved_aesthetics_6.5plus_with_images), and then fine-tuning was performed on the [LAION HighRes data](https://huggingface.co/datasets/laion/laion-high-resolution).

The main Text2Image diffusion model was trained on the basis of 170M text-image pairs from the [LAION HighRes dataset](https://huggingface.co/datasets/laion/laion-high-resolution) (an important condition was the presence of images with a resolution of at least 768x768). The use of 170M pairs is due to the fact that we kept the UNet diffusion block from Kandinsky 2.0, which allowed us not to train it from scratch. Further, at the stage of fine-tuning, a dataset of 2M very high-quality high-resolution images with descriptions (COYO, anime, landmarks_russia, and a number of others) was used separately collected from open sources.


### Evaluation
We quantitatively measure the performance of Kandinsky 2.1 on the COCO_30k dataset, in zero-shot mode. The table below presents FID.

FID metric values ​​for generative models on COCO_30k
|    | FID (30k)|
|:------|----:|
| eDiff-I (2022) | 6.95 | 
| Image (2022) | 7.27 | 
| Kandinsky 2.1 (2023) | 8.21|
| Stable Diffusion 2.1 (2022) | 8.59 | 
| GigaGAN, 512x512 (2023) | 9.09 | 
| DALL-E 2 (2022) | 10.39 | 
| GLIDE (2022) | 12.24 | 
| Kandinsky 1.0 (2022) | 15.40 | 
| DALL-E (2021) | 17.89 | 
| Kandinsky 2.0 (2022) | 20.00 | 
| GLIGEN (2022) | 21.04 | 

For more information, please refer to the upcoming technical report.

## BibTex
If you find this repository useful in your research, please cite:
```
@misc{kandinsky 2.1,
  title         = {kandinsky 2.1},
  author        = {Arseniy Shakhmatov, Anton Razzhigaev, Aleksandr Nikolich, Vladimir Arkhipkin, Igor Pavlov, Andrey Kuznetsov, Denis Dimitrov},
  year          = {2023},
  howpublished  = {},
}
```