pinecoresystems commited on
Commit
48c6fb4
·
verified ·
1 Parent(s): a154878

Upload UPSTREAM-README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. UPSTREAM-README.md +96 -96
UPSTREAM-README.md CHANGED
@@ -1,96 +1,96 @@
1
- ---
2
- license: mit
3
- library_name: custom
4
- pipeline_tag: image-text-to-image
5
- inference: false
6
- tags:
7
- - image-text-to-image
8
- - layer-decomposition
9
- - rgba
10
- - graphic-design
11
- ---
12
-
13
- # Ming-Image-0.1-Design-Layer
14
-
15
- [🧩 ModelScope](https://www.modelscope.cn/models/inclusionAI/Ming-Image-0.1-Design-Layer) · [🤗 Hugging Face](https://huggingface.co/inclusionAI/Ming-Image-0.1-Design-Layer) · [📄 Blog](https://mp.weixin.qq.com/s/VGdtxfM8kbHIQJw50VD_Sw) · [🖥️ Demo](https://huggingface.co/spaces/Xiaolong-Wang/Ming-Image-0.1-Design-Layer)<br>
16
- [🎨 Design Skill](https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/ling-ui-design) · [📊 PPT Skill](https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/image-to-editable-ppt)
17
-
18
- Ming-Image-0.1-Design-Layer decomposes a flattened design image into a
19
- requested number of RGBA layers using an image and a layer plan.
20
-
21
- ## Quick Start
22
-
23
- Use the companion [Ming-Image repository](https://github.com/inclusionAI/Ming-Image)
24
- for installation and inference:
25
-
26
- ```bash
27
- git clone https://github.com/inclusionAI/Ming-Image
28
- cd Ming-Image
29
- pip install -r requirements.txt
30
-
31
- python infer.py \
32
- --model inclusionAI/Ming-Image-0.1-Design-Layer \
33
- --task layer-decompose \
34
- --input-image assets/layer_samples/card_making_input.png \
35
- --prompt assets/layer_samples/card_making_prompt.txt \
36
- --attn-implementation flash_attention_2 \
37
- --resolution 1024 \
38
- --output-dir outputs/layers
39
- ```
40
-
41
- This runs the released six-layer card-making example. See the full
42
- [layer-decomposition demo](https://github.com/inclusionAI/Ming-Image#layer-decomposition-demo)
43
- for the output structure and layer-count behavior.
44
-
45
- Prompt enhancement (PE) can use `Ling-3.0-flash-VL` or `qwen3.8-27B`; see
46
- [layer-decomposition prompt rewriting](https://github.com/inclusionAI/Ming-Image#layer-decomposition-prompt-rewriting).
47
-
48
- ## Deployment
49
-
50
- We recommend the following inference frameworks to serve the model:
51
-
52
- - vLLM-Omni: see the [recipes](https://github.com/vllm-project/vllm-omni/blob/main/recipes/inclusionAI/Ming-Image.md)
53
- and [installation guide](https://docs.vllm.ai/projects/vllm-omni/en/latest/getting_started/quickstart/).
54
-
55
- ## Recommended settings
56
-
57
- - Working-resolution bucket: **1024** (recommended), or **512** for faster
58
- layer decomposition. The output preserves the input image's aspect ratio.
59
- - Sampling steps: **12**.
60
- - CFG scale: **2.0**.
61
- - Precision: **BF16**.
62
- - Hardware: **one CUDA GPU with 80 GiB VRAM** (validated configuration).
63
-
64
- Provide `--input-image` plus either a detailed layer specification through
65
- `--prompt`, or omit `--prompt` and set `--num-layers N` to create the default
66
- request. When `--prompt` is supplied, the layer count declared in that prompt
67
- controls the output count. The standalone outputs are saved as RGBA PNG files.
68
-
69
- ## Gallery
70
-
71
- <p align="center">
72
- <img src="./assets/showcase.webp" width="100%" alt="Ming-Image-0.1-Design-Layer decomposition example">
73
- </p>
74
-
75
- The example shows the input design, six decomposed layers, and the recomposed
76
- result.
77
-
78
- <p align="center">
79
- <img src="./assets/gallery.webp" width="100%" alt="Ming-Image-0.1-Design-Layer gallery">
80
- </p>
81
-
82
- The gallery shows additional flattened designs, their transparent layers, and
83
- the corresponding recomposed results.
84
-
85
- ## Performance
86
-
87
- <p align="center">
88
- <img src="./assets/performance.webp" width="100%" alt="Layer-decomposition results on the Crello test set">
89
- </p>
90
-
91
- The table reports quantitative layer-decomposition results on the Crello test
92
- set; lower RGB L1 and higher Alpha soft IoU are better.
93
-
94
- ## License
95
-
96
- This model is released under the [MIT License](./LICENSE).
 
1
+ ---
2
+ license: mit
3
+ library_name: custom
4
+ pipeline_tag: image-text-to-image
5
+ inference: false
6
+ tags:
7
+ - image-text-to-image
8
+ - layer-decomposition
9
+ - rgba
10
+ - graphic-design
11
+ ---
12
+
13
+ # Ming-Image-0.1-Design-Layer
14
+
15
+ [🧩 ModelScope](https://www.modelscope.cn/models/inclusionAI/Ming-Image-0.1-Design-Layer) · [🤗 Hugging Face](https://huggingface.co/inclusionAI/Ming-Image-0.1-Design-Layer) · [📄 Blog](https://mp.weixin.qq.com/s/VGdtxfM8kbHIQJw50VD_Sw) · [🖥️ Demo](https://huggingface.co/spaces/Xiaolong-Wang/Ming-Image-0.1-Design-Layer)<br>
16
+ [🎨 Design Skill](https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/ling-ui-design) · [📊 PPT Skill](https://github.com/inclusionAI/ling-cookbook/tree/main/resources/recommended-skills/image-to-editable-ppt)
17
+
18
+ Ming-Image-0.1-Design-Layer decomposes a flattened design image into a
19
+ requested number of RGBA layers using an image and a layer plan.
20
+
21
+ ## Quick Start
22
+
23
+ Use the companion [Ming-Image repository](https://github.com/inclusionAI/Ming-Image)
24
+ for installation and inference:
25
+
26
+ ```bash
27
+ git clone https://github.com/inclusionAI/Ming-Image
28
+ cd Ming-Image
29
+ pip install -r requirements.txt
30
+
31
+ python infer.py \
32
+ --model inclusionAI/Ming-Image-0.1-Design-Layer \
33
+ --task layer-decompose \
34
+ --input-image assets/layer_samples/card_making_input.png \
35
+ --prompt assets/layer_samples/card_making_prompt.txt \
36
+ --attn-implementation flash_attention_2 \
37
+ --resolution 1024 \
38
+ --output-dir outputs/layers
39
+ ```
40
+
41
+ This runs the released six-layer card-making example. See the full
42
+ [layer-decomposition demo](https://github.com/inclusionAI/Ming-Image#layer-decomposition-demo)
43
+ for the output structure and layer-count behavior.
44
+
45
+ Prompt enhancement (PE) can use `Ling-3.0-flash-VL` or `qwen3.8-27B`; see
46
+ [layer-decomposition prompt rewriting](https://github.com/inclusionAI/Ming-Image#layer-decomposition-prompt-rewriting).
47
+
48
+ ## Deployment
49
+
50
+ We recommend the following inference frameworks to serve the model:
51
+
52
+ - vLLM-Omni: see the [recipes](https://github.com/vllm-project/vllm-omni/blob/main/recipes/inclusionAI/Ming-Image.md)
53
+ and [installation guide](https://docs.vllm.ai/projects/vllm-omni/en/latest/getting_started/quickstart/).
54
+
55
+ ## Recommended settings
56
+
57
+ - Working-resolution bucket: **1024** (recommended), or **512** for faster
58
+ layer decomposition. The output preserves the input image's aspect ratio.
59
+ - Sampling steps: **12**.
60
+ - CFG scale: **2.0**.
61
+ - Precision: **BF16**.
62
+ - Hardware: **one CUDA GPU with 80 GiB VRAM** (validated configuration).
63
+
64
+ Provide `--input-image` plus either a detailed layer specification through
65
+ `--prompt`, or omit `--prompt` and set `--num-layers N` to create the default
66
+ request. When `--prompt` is supplied, the layer count declared in that prompt
67
+ controls the output count. The standalone outputs are saved as RGBA PNG files.
68
+
69
+ ## Gallery
70
+
71
+ <p align="center">
72
+ <img src="./assets/showcase.webp" width="100%" alt="Ming-Image-0.1-Design-Layer decomposition example">
73
+ </p>
74
+
75
+ The example shows the input design, six decomposed layers, and the recomposed
76
+ result.
77
+
78
+ <p align="center">
79
+ <img src="./assets/gallery.webp" width="100%" alt="Ming-Image-0.1-Design-Layer gallery">
80
+ </p>
81
+
82
+ The gallery shows additional flattened designs, their transparent layers, and
83
+ the corresponding recomposed results.
84
+
85
+ ## Performance
86
+
87
+ <p align="center">
88
+ <img src="./assets/performance.webp" width="100%" alt="Layer-decomposition results on the Crello test set">
89
+ </p>
90
+
91
+ The table reports quantitative layer-decomposition results on the Crello test
92
+ set; lower RGB L1 and higher Alpha soft IoU are better.
93
+
94
+ ## License
95
+
96
+ This model is released under the [MIT License](./LICENSE).