Text2Human: Text-Driven Controllable Human Image Generation

  • 1S-Lab, Nanyang Technological University
  • 2SenseTime Research
ACM Transactions on Graphics (Proceedings of SIGGRAPH), 2022

Generating high-quality and diverse human images is an important yet challenging task in vision and graphics. However, existing generative models often fall short under the high diversity of clothing shapes and textures. Furthermore, the generation process is even desired to be intuitively controllable for layman users. In this work, we present a text-driven controllable framework, Text2Human, for a high-quality and diverse human generation. We synthesize full-body human images starting from a given human pose with two dedicated steps. 1) With some texts describing the shapes of clothes, the given human pose is first translated to a human parsing map. 2) The final human image is then generated by providing the system with more attributes about the textures of clothes. Specifically, to model the diversity of clothing textures, we build a hierarchical texture-aware codebook that stores multi-scale neural representations for each type of texture. The codebook at the coarse level includes the structural representations of textures, while the codebook at the fine level focuses on the details of textures. To make use of the learned hierarchical codebook to synthesize desired images, a diffusionbased transformer sampler with mixture of experts is firstly employed to sample indices from the coarsest level of the codebook, which then is used to predict the indices of the codebook at finer levels. The predicted indices at different levels are translated to human images by the decoder learned accompanied with hierarchical codebooks. The use of mixture-of-experts allows for the generated image conditioned on the fine-grained text input. The prediction for finer level indices refines the quality of clothing textures. Extensive quantitative and qualitative evaluations demonstrate that our proposed Text2Human framework can generate more diverse and realistic human images compared to state-of-the-art methods.

You can select the attributes to customize the synthesized human images.

A wearing a with , and with .
A wearing a with , and with .
A lady wearing a with , and with .
A lady wearing a with , and with .

DeepFashion-MultiModal is a large-scale high-quality human dataset with rich multi-modal annotations. It has the following properties:

1. It contains 44,096 high-resolution human images, including 12,701 full body human images.

2. For each full body images, we manually annotate the human parsing labels of 24 classes.

3. For each full body images, we manually annotate the keypoints.

4. We extract DensePose for each human image.

5. Each image is manually annotated with attributes for both clothes shapes and textures.

6. We provide a textual description for each image.

        title={Text2Human: Text-Driven Controllable Human Image Generation},
        author={Jiang, Yuming and Yang, Shuai and Qiu, Haonan and Wu, Wayne and Loy, Chen Change and Liu, Ziwei},
        journal={ACM Transactions on Graphics (TOG)},
        publisher={ACM New York, NY, USA},
More Researches

EVA3D proposes a compositional framework to generate 3D human from 2D image collections.

StyleGAN-Human investigates the "data engineering" in unconditional human generation.

Talk-to-Edit proposes a StyleGAN-based method and a multi-modal dataset for dialog-based facial editing.

CelebA-Dialog is a large-scale visual-language face dataset with rich fine-grained labels and textual descriptions.

DualStyleGAN proposes StyleGAN-based Cartoonization.

AvatarCLIP proposes a zero-shot text-driven framework for 3D avatar generation and animation

For more interesting researches, please visit the MMLab@NTU website.

We referred to the project page of Nerfies and AvatarCLIP when creating this project page.