Update README.md

README.md

@@ -34,16 +34,26 @@ The control ability is also strong, for example if you are unstatisfied with som
 
 - **Paper [optional]:** https://arxiv.org/abs/2302.05543
 
-### Examples
+### Examples[**Note the following examples are all generate using stabilityai/stable-diffusion-xl-base-1.0 and xinsir/controlnet-scribble-sdxl-1.0**]
+prompt: purple feathered eagle with specks of light like stars in feathers. It glows with arcane power
 ![image1](./000155_scribble_concat.webp)
+prompt: manga girl in the city, drip marketing
 ![image2](./000186_scribble_concat.webp)
+prompt: 17 year old girl with long dark hair in the style of realism with fantasy elements, detailed botanical illustrations, barbs and thorns, ethereal, magical, black, purple and maroon, intricate, photorealistic 
 ![image3](./000210_scribble_concat.webp)
+prompt: a logo for a paintball field named district 7 on a white background featuring paintballs the is bright and colourful eye catching and impactuful
 ![image4](./000227_scribble_concat.webp)
+prompt: a photograph of a handsome crying blonde man with his face painted in the pride flag 
 ![image5](./000242_scribble_concat.webp)
+prompt: simple flat sketch fox play ball 
 ![image6](./000250_scribble_concat.webp)
+prompt: concept art, a surreal magical Tome of the Sun God, the book binding appears to be made of solar fire and emits a holy, radiant glow, Age of Wonders, Unreal Engine v5
 ![image7](./000256_scribble_concat.webp)
+prompt: black Caribbean man walking balance front his fate chaos anarchy liberty independence force energy independence cinematic surreal beautiful rendition intricate sharp detail 8k 
 ![image8](./000271_scribble_concat.webp)
+prompt: die hard nakatomi plaza, explosion at the top, vector, night scene 
 ![image9](./000285_scribble_concat.webp)
+prompt: solitary glowing yellow tree in a desert. ultra wide shot. night time. hdr photography 
 ![image10](./000290_scribble_concat.webp)
 
 
@@ -157,3 +167,28 @@ images = pipe(
 images[0].save(f"your image save path, png format is usually better than jpg or webp in terms of image quality but got much bigger")
 ```
 
+## Evaluation Data
+The test data is randomly sample from midjourney upscale images with prompts, as the purpose of the project is to letting people draw images like midjourney. midjourney’s users include a large number of professional designers,
+and the upscale image tend to have more beauty score and prompt consistency, it is suitable to use it as the test set to judge the ability of controlnet. We select 300 prompt-image pairs randomly and generate 4 images per prompt,
+totally 1200 images generated. We caculate the Laion Aesthetic Score to measure the beauty and the PerceptualSimilarity to measure the control ability, we find the quality of images have a good consistency with the meric values.
+We compare our methods with other SOTA huggingface models and list the result below. We are the models that have highest aesthectic score, and can generate visually appealing images if you prompt it properly.
+
+Note: The condition image are generated using HED detector and random threshold to generate different kinds of lines.
+
+## Quantitative Result
+| metric | xinsir/controlnet-scribble-sdxl-1.0 | 
+|-------|-------|
+| laion_aesthetic |  **6.03**   |
+| perceptual similarity | 0.5701 |
+
+laion_aesthetic(the higher the better)  
+perceptual similarity(the lower the better)
+
+Note: The values are caculated when save in webp format, when save in png the aesthetic values will increase 0.1-0.3, but the relative relation remains unchanged.
+
+### Conclusion
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+In our evaluation, the model can generate visually appealing images using simple sketch and simple prompt. This model can support any type of lines and any width of lines, using thick line will give a coarse control
+which obey the prompt your write more, and using thick line will give a strong control which obey the condition image more. The model can help you complish the drawing from coarse to fine, the model achieves higher 
+aesthetic score than xinsir/controlnet-canny-sdxl-1.0, but the control ability will decrease a bit because of thick line.