SimpModeling: Sketching Implicit Field to Guide Mesh Modeling for 3D Animalmorphic Head Design


Zhongjin Luo1      Jie Zhou2      Heming Zhu1      Dong Du1      Xiaoguang Han1*      Hongbo Fu2*     

*Corresponding email: hanxiaoguang@cuhk.edu.cn; hongbofu@cityu.edu.hk

1SSE, The Chinese University of Hong Kong, Shenzhen       2School of Creative Media, City University of Hong Kong      

If you are interested in sketch-based 3D modeling, you can also refer to SketchMetaFace and Sketch2RaBit.



Figure 1. We present SimpModeling, a novel sketching system designed for amateur users to create desired animalmorphic heads. It provides two stages for mesh modeling: coarse shape sketching where users may create coarse head models with 3D curve handles (blue), and geometric detail crafting where users may add geometric surface details by drawing sketches (red) on the coarse models. The two animalmorphic head models in this figure were created by a novice user without any 3D modeling experiences in ten minutes.

Introduction

Head shapes play an important role in 3D character design. In this work, we propose SimpModeling, a novel sketch-based system for helping users, especially amateur users, easily model 3D animalmorphic heads - a prevalent kind of heads in character design. Although sketching provides an easy way to depict desired shapes, it is challenging to infer dense geometric information from sparse line drawings. Recently, deepnet-based approaches have been taken to address this challenge and try to produce rich geometric details from very few strokes. However, while such methods reduce users' workload, they would cause less controllability of target shapes. This is mainly due to the uncertainty of the neural prediction. Our system tackles this issue and provides good controllability from three aspects: 1) we separate coarse shape design and geometric detail specification into two stages and respectively provide different sketching means; 2) in coarse shape designing, sketches are used for both shape inference and geometric constraints to determine global geometry, and in geometric detail crafting, sketches are used for carving surface details; 3) in both stages, we use the advanced implicit-based shape inference methods, which have strong ability to handle the domain gap between freehand sketches and synthetic ones used for training. Experimental results confirm the effectiveness of our method and the usability of our interactive system. We also contribute to a dataset of high-quality 3D animal heads, which are manually created by artists.

The major contributions of this work are summarized as follows:

Demo


Publication

Paper - ArXiv - pdf (abs) | Code

If you find our work useful, please consider citing it:


@inproceedings{luo2021simpmodeling,
    title={Simpmodeling: Sketching implicit field to guide mesh modeling for 3d animalmorphic head design},
    author={Luo, Zhongjin and Zhou, Jie and Zhu, Heming and Du, Dong and Han, Xiaoguang and Fu, Hongbo},
    booktitle={The 34th Annual ACM Symposium on User Interface Software and Technology},
    pages={854--863},
    year={2021}
}
        

Dataset

We propose 3DAnimalHead, the largest animalmorphic head dataset by far. Our dataset contains 1,955 high-quality animalmorphic head models manually created by artists, covering 17 common categories. Each animalmorphic head is carefully annotated with contour curves to facilitate future research on 3D animalmorphic head generation.

Data Examples

Figure 2. Examples of 3DAnimalHead.

Dataset Statistics

Table 1 shows the dataset statistics for SAniHead and our proposed 3DAnimalHead. As seen, 3DAnimalHead is one scale larger than SAniHead and enjoys much better data balance across categories.


Table 1. Dataset statistics for SAniHead and our proposed 3DAnimalHead.

Evaluation

Usability Study

Since our proposed system is designed to assist users in instantiating their ideas, in the usability study, we asked the participants to create whatever they wanted with our system within 30 minutes and did not ask them to create a specific number of models. We observed that each participant created on average 3 to 4 models within the given time. Figure 3 shows representative models created by the subjects with our system. It can be seen from this figure that our system supports users with limited knowledge in geometrical modeling to create animalmorphic heads with diversified shapes and rich geometric details.


Figure 3. The gallery of the representative models created using our system by 15 invited amateur users.

Comparison Study

After thoroughly reviewing the existing open-source sketch-based modeling systems, we choose FiberMesh and SAniHead, which are most relevant to our work, for comparison. In the Comparison Study, each user was given different reference images and asked to create 3D animalmorphic heads according to the reference image with three systems(our system, FiberMesh, and SAniHead) in a randomized order. For each system, the participants were given at most 15 minutes. Note that the created 3D model was not required to strictly follow the reference image and differences were allowed. Figure 4 shows the animalmorphic head models created using the compared three systems.


Figure 4. The models created in the comparison study.

Results Gallery

Figure 5 are sampled results created by amateur users using our system. As seen, given limited time, novice users working with our system can create models with visually appealing shapes and vivid details easily with a few sketch strokes.


Figure 5. Results created by amateur users using our system.