dataautogpt3 commited on
Commit
8a9ccc8
1 Parent(s): 82221ee

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -80
README.md CHANGED
@@ -1,80 +1 @@
1
- ---
2
- pipeline_tag: text-to-image
3
- widget:
4
- - text: >-
5
- movie scene screencap, cinematic footage. thanos smelling a little yellow
6
- rose. extreme wide angle,
7
- output:
8
- url: 1man.png
9
- - text: 'A tiny robot taking a break under a tree in the garden '
10
- output:
11
- url: robot.png
12
- - text: mystery
13
- output:
14
- url: mystery.png
15
- - text: a cat wearing sunglasses in the summer
16
- output:
17
- url: cat.png
18
- - text: 'robot holding a sign that says ’a storm is coming’ '
19
- output:
20
- url: storm.png
21
- - text: >-
22
- The Exegenesis of the soul, captured within a boundless well of starlight,
23
- pulsating and vibrating wisps, chiaroscuro, humming transformer
24
- output:
25
- url: soul.png
26
- - text: >-
27
- Lady of War, chique dark clothes, vinyl, imposing pose, anime style, 90s
28
- natural photography of a man, glasses, cinematic,
29
- output:
30
- url: anime.png
31
- - text: natural photography of a man, glasses, cinematic,
32
- output:
33
- url: glasses.png
34
- - text: if I could turn back time
35
- output:
36
- url: time.png
37
- - text: god
38
- output:
39
- url: god.png
40
- - text: >-
41
- cineamantic, aesthetic, best quality, masterpiece, powerful aura, fog, text
42
- logo, "Mobius"
43
- output:
44
- url: mobius.png
45
- - text: the backrooms
46
- output:
47
- url: backrooms.png
48
- license: apache-2.0
49
- ---
50
- <Gallery />
51
-
52
- # Mobius: Redefining State-of-the-Art in Debiased Diffusion Models
53
-
54
- Mobius, a revolutionary diffusion model, pushes the boundaries of domain-agnostic debiasing and representation realignment. By employing the cutting-edge constructive deconstruction framework, Mobius achieves unrivaled generalization across a vast array of styles and domains, eliminating the need for expensive pretraining from scratch.
55
-
56
- # Domain-Agnostic Debiasing: A Groundbreaking Approach
57
-
58
- Domain-agnostic debiasing is a novel technique pioneered Corcel. This innovative approach aims to remove biases inherent in diffusion models without limiting their ability to generalize across diverse domains. Traditional debiasing methods often focus on specific domains or styles, resulting in models that struggle to adapt to new or unseen contexts. In contrast, domain-agnostic debiasing ensures that the model remains unbiased while maintaining its versatility and adaptability.
59
-
60
- The key to domain-agnostic debiasing lies in the constructive deconstruction framework, a proprietary method developed by the Corcel Diffusion team. This framework allows for fine-grained reworking of biases and representations without the need for pretraining from scratch. The technical details of this groundbreaking approach will be discussed in an upcoming research paper, "Constructive Deconstruction: Domain-Agnostic Debiasing of Diffusion Models," which will be made available on the Corcel.io website and through scientific publications.
61
-
62
- By applying domain-agnostic debiasing, Mobius sets a new standard for fairness and impartiality in image generation while maintaining its exceptional ability to adapt to a wide range of styles and domains.
63
-
64
- # Surpassing the State-of-the-Art
65
-
66
- Mobius outperforms existing state-of-the-art diffusion models in several key areas:
67
-
68
- Unbiased generation: Mobius generates images that are virtually free from the inherent biases commonly found in other diffusion models, setting a new benchmark for fairness and impartiality across all domains.
69
-
70
- Exceptional generalization: With its unparalleled ability to adapt to an extensive range of styles and domains, Mobius consistently delivers top-quality results, surpassing the limitations of previous models.
71
-
72
- Efficient fine-tuning: The Mobius base model serves as a superior foundation for creating specialized models tailored to specific tasks or domains, requiring significantly less fine-tuning and computational resources compared to other state-of-the-art models.
73
-
74
-
75
- # Usage and Recommendations
76
-
77
- - a CFG of either 3.5 to 7
78
- - Requires a CLIP skip of -3
79
- - highly suggested to preappenmed watermark to all negatives and keep negatives simple such as "watermark" or "worst, watermark"
80
-
 
1
+ Coming Soon