Add some images
Browse files
README.md
CHANGED
|
@@ -55,16 +55,18 @@ image = pipe(prompt=prompt).images[0]
|
|
| 55 |
|
| 56 |
### User Study
|
| 57 |
|
|
|
|
|
|
|
| 58 |
According to user studies conducted by Playground, involving over 2,600 prompts and thousands of users, the images generated by Playground v2 are favored 2.5 times more than those produced by [Stable Diffusion XL](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0).
|
| 59 |
|
| 60 |
We report user preference metrics on [PartiPrompts](https://github.com/google-research/parti), following standard practice, and on an internal prompt dataset curated by the Playground team. The “Internal 1K” prompt dataset is diverse and covers various categories and tasks.
|
| 61 |
|
| 62 |
During the user study, we give users instructions to evaluate image pairs based on both (1) their aesthetic preference and (2) the image-text alignment.
|
| 63 |
|
| 64 |
-
< [Aleks] show a screenshot of the Playground UI here? [Daiqing] I think it is fine to just explain>
|
| 65 |
-
|
| 66 |
### MJHQ-30K Benchmark
|
| 67 |
|
|
|
|
|
|
|
| 68 |
We introduce a new benchmark, [MJHQ-30K](https://huggingface.co/datasets/playgroundai/MJHQ30K), for automatic evaluation of a model’s aesthetic quality. The benchmark computes FID on a high-quality dataset to gauge aesthetic quality.
|
| 69 |
|
| 70 |
We curate the high-quality dataset from Midjourney with 10 common categories, each category with 3K samples. Following common practice, we use aesthetic score and CLIP score to ensure high image quality and high image-text alignment. Furthermore, we take extra care to make the data diverse within each category.
|
|
|
|
| 55 |
|
| 56 |
### User Study
|
| 57 |
|
| 58 |
+
|
| 59 |
+
|
| 60 |
According to user studies conducted by Playground, involving over 2,600 prompts and thousands of users, the images generated by Playground v2 are favored 2.5 times more than those produced by [Stable Diffusion XL](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0).
|
| 61 |
|
| 62 |
We report user preference metrics on [PartiPrompts](https://github.com/google-research/parti), following standard practice, and on an internal prompt dataset curated by the Playground team. The “Internal 1K” prompt dataset is diverse and covers various categories and tasks.
|
| 63 |
|
| 64 |
During the user study, we give users instructions to evaluate image pairs based on both (1) their aesthetic preference and (2) the image-text alignment.
|
| 65 |
|
|
|
|
|
|
|
| 66 |
### MJHQ-30K Benchmark
|
| 67 |
|
| 68 |
+
|
| 69 |
+
|
| 70 |
We introduce a new benchmark, [MJHQ-30K](https://huggingface.co/datasets/playgroundai/MJHQ30K), for automatic evaluation of a model’s aesthetic quality. The benchmark computes FID on a high-quality dataset to gauge aesthetic quality.
|
| 71 |
|
| 72 |
We curate the high-quality dataset from Midjourney with 10 common categories, each category with 3K samples. Following common practice, we use aesthetic score and CLIP score to ensure high image quality and high image-text alignment. Furthermore, we take extra care to make the data diverse within each category.
|