@@ -157,7 +166,7 @@

- Recent Articles + Recent Researches

@@ -175,18 +184,18 @@

- 《Kolors: Unet is enough for diffusion》 + Kolors: Unet is enough for diffusion (arXiv-2024)

read more @@ -206,19 +215,18 @@

- 《Learning Multi-dimensional Human Preference for Text-to-Image Generation》,CVPR-2024 + Learning Multi-dimensional Human Preference for Text-to-Image Generation (CVPR-2024)

read more @@ -252,6 +260,15 @@

+ + + + + + + + + diff --git a/index.xml b/index.xml index d8256d2..4a5bf5d 100644 --- a/index.xml +++ b/index.xml @@ -9,32 +9,32 @@ Fri, 21 Jun 2024 11:00:59 -0400 - 《Kolors: Unet is enough for diffusion》 + Kolors: Unet is enough for diffusion (arXiv-2024) https://Kwai-Kolors.github.io/post/post-2/ Fri, 21 Jun 2024 11:00:59 -0400 https://Kwai-Kolors.github.io/post/post-2/ - Table of Contents 📖 Model Introduction 📊 Evaluation Performance 🥇🥇🔥🔥 🎥 Visualization 🛠️ Quick Start 📜 License and Citation 📖 Model Introduction We have open-sourced the Kolors large model, which is a large-scale text-to-image generation model based on latent diffusion. The current open-source model has 2.7 billion parameters. Kolors is trained on billions of text-image pairs and demonstrates significant advantages in visual quality, complex semantic understanding, and even text generation (Chinese and English characters) compared to both open-source and closed-source models. + We have open-sourced the <strong>Kolors</strong> large model, which is a large-scale text-to-image generation model based on latent diffusion. The current open-source model has 2.7 billion parameters. Kolors is trained on billions of text-image pairs and demonstrates significant advantages in visual quality, complex semantic understanding, and even text generation (Chinese and English characters) compared to both open-source and closed-source models. Additionally, Kolors supports both Chinese and English, making it more competitive in understanding Chinese-specific content. The model link is provided in the table below; please click to learn more. - 《Learning Multi-dimensional Human Preference for Text-to-Image Generation》,CVPR-2024 + Learning Multi-dimensional Human Preference for Text-to-Image Generation (CVPR-2024) https://Kwai-Kolors.github.io/post/post-1/ Thu, 20 Jun 2024 10:58:08 -0400 https://Kwai-Kolors.github.io/post/post-1/ - Authors Sixian Zhang, Bohan Wang, Junqiang Wu*, Yan Li‡, Tingting Gao, Di Zhang, Zhongyuan Wang Links Paper Code arXiv Abstract Current metrics for text-to-image models typically rely on statistical metrics which inadequately represent the real preference of humans. Although recent works attempt to learn these preferences via human annotated images, they reduce the rich tapestry of human preference to a single overall score. However, the preference results vary when humans evaluate images with different aspects. + To learn the multi-dimensional human preferences, we propose the Multi-dimensional Preference Score (MPS), the first multi-dimensional preference scoring model for the evaluation of text-to-image models. The MPS introduces the preference condition module upon CLIP model to learn these diverse preferences. It is trained based on our Multi-dimensional Human Preference (MHP) Dataset, which comprises 918,315 human preference choices across 4 dimensions on 607,541 images. The images are generated by a wide range of latest text-to-image models. The MPS outperforms existing scoring methods across 3 datasets in 4 dimensions, enabling it a promising metric for evaluating and improving text-to-image generation. About Us https://Kwai-Kolors.github.io/about/ Mon, 01 Jan 0001 00:00:00 +0000 https://Kwai-Kolors.github.io/about/ - KOLORS Team is an innovative team focused on the research and development of visual generation and editing algorithms. We are dedicated to advancing the forefront of visual generation technology and applying the latest research findings to practical business scenarios. + Kolors Team is an innovative group focused on the research and development of visual generation and editing algorithms. We are dedicated to advancing the forefront of visual generation technology and applying the latest research findings to practical business scenarios. Contact https://Kwai-Kolors.github.io/contact/ Mon, 01 Jan 0001 00:00:00 +0000 https://Kwai-Kolors.github.io/contact/ - If you encounter any issues while using KOLORS, feel free to contact us via the email kwai-kolors@kuaishou.com. Your Name Email Address An email address is required. Message + If you encounter any issues while using Kolors, feel free to contact us via email kwai-kolors@kuaishou.com. Your Name Email Address An email address is required. Message diff --git a/post/8_up.png b/post/8_up.png new file mode 100644 index 0000000..131edfc Binary files /dev/null and b/post/8_up.png differ diff --git a/post/index.html b/post/index.html index ddfa2a6..d500249 100644 --- a/post/index.html +++ b/post/index.html @@ -4,7 +4,7 @@ - Articles | Kolors + Researches | Kolors @@ -42,18 +42,18 @@ - + - + - + @@ -90,14 +90,14 @@
  • - - Articles + + Contact
  • - - Contact + + Researches
  • @@ -113,6 +113,15 @@ + + + + + + + + + @@ -126,7 +135,7 @@

    - Articles + Researches

    @@ -139,8 +148,7 @@

    - -

    We will publish the team’s latest papers and models on this page. The page will be continuously updated.

    +

    We will publish the team’s latest papers and models on this page. The page will be continuously updated.

    @@ -173,12 +181,11 @@

    - 《Learning Multi-dimensional Human Preference for Text-to-Image Generation》,CVPR-2024 + Learning Multi-dimensional Human Preference for Text-to-Image Generation (CVPR-2024)

    @@ -204,6 +211,15 @@

    + + + + + + + + + diff --git a/post/index.xml b/post/index.xml index 9faa998..7fbb585 100644 --- a/post/index.xml +++ b/post/index.xml @@ -1,26 +1,26 @@ - Articles on Kolors + Researches on Kolors https://Kwai-Kolors.github.io/post/ - Recent content in Articles on Kolors + Recent content in Researches on Kolors Hugo en Fri, 21 Jun 2024 11:00:59 -0400 - 《Kolors: Unet is enough for diffusion》 + Kolors: Unet is enough for diffusion (arXiv-2024) https://Kwai-Kolors.github.io/post/post-2/ Fri, 21 Jun 2024 11:00:59 -0400 https://Kwai-Kolors.github.io/post/post-2/ - Table of Contents 📖 Model Introduction 📊 Evaluation Performance 🥇🥇🔥🔥 🎥 Visualization 🛠️ Quick Start 📜 License and Citation 📖 Model Introduction We have open-sourced the Kolors large model, which is a large-scale text-to-image generation model based on latent diffusion. The current open-source model has 2.7 billion parameters. Kolors is trained on billions of text-image pairs and demonstrates significant advantages in visual quality, complex semantic understanding, and even text generation (Chinese and English characters) compared to both open-source and closed-source models. + We have open-sourced the <strong>Kolors</strong> large model, which is a large-scale text-to-image generation model based on latent diffusion. The current open-source model has 2.7 billion parameters. Kolors is trained on billions of text-image pairs and demonstrates significant advantages in visual quality, complex semantic understanding, and even text generation (Chinese and English characters) compared to both open-source and closed-source models. Additionally, Kolors supports both Chinese and English, making it more competitive in understanding Chinese-specific content. The model link is provided in the table below; please click to learn more. - 《Learning Multi-dimensional Human Preference for Text-to-Image Generation》,CVPR-2024 + Learning Multi-dimensional Human Preference for Text-to-Image Generation (CVPR-2024) https://Kwai-Kolors.github.io/post/post-1/ Thu, 20 Jun 2024 10:58:08 -0400 https://Kwai-Kolors.github.io/post/post-1/ - Authors Sixian Zhang, Bohan Wang, Junqiang Wu*, Yan Li‡, Tingting Gao, Di Zhang, Zhongyuan Wang Links Paper Code arXiv Abstract Current metrics for text-to-image models typically rely on statistical metrics which inadequately represent the real preference of humans. Although recent works attempt to learn these preferences via human annotated images, they reduce the rich tapestry of human preference to a single overall score. However, the preference results vary when humans evaluate images with different aspects. + To learn the multi-dimensional human preferences, we propose the Multi-dimensional Preference Score (MPS), the first multi-dimensional preference scoring model for the evaluation of text-to-image models. The MPS introduces the preference condition module upon CLIP model to learn these diverse preferences. It is trained based on our Multi-dimensional Human Preference (MHP) Dataset, which comprises 918,315 human preference choices across 4 dimensions on 607,541 images. The images are generated by a wide range of latest text-to-image models. The MPS outperforms existing scoring methods across 3 datasets in 4 dimensions, enabling it a promising metric for evaluating and improving text-to-image generation. diff --git a/post/post-1/index.html b/post/post-1/index.html index 895ae56..19716d7 100644 --- a/post/post-1/index.html +++ b/post/post-1/index.html @@ -4,7 +4,7 @@ - 《Learning Multi-dimensional Human Preference for Text-to-Image Generation》,CVPR-2024 | Kolors + Learning Multi-dimensional Human Preference for Text-to-Image Generation (CVPR-2024) | Kolors @@ -37,7 +37,7 @@ - + @@ -46,14 +46,14 @@ - + - + @@ -70,7 +70,7 @@ -
    +
    @@ -199,11 +207,11 @@

    - 《Kolors: Unet is enough for diffusion》 + Kolors: Unet is enough for diffusion (arXiv-2024)