Computer Graphics Forum最新文献

Automatic font generation aims to streamline the design process by creating new fonts with minimal style references. This technology significantly reduces the manual labour and costs associated with traditional font design. Image-to-image translation has been the dominant approach, transforming font images from a source style to a target style using a few reference images. However, this framework struggles to fully decouple content from style, particularly when dealing with significant style shifts. Despite these limitations, image-to-image translation remains prevalent due to two main challenges faced by conditional generative models: (1) inability to handle unseen characters and (2) difficulty in providing precise content representations equivalent to the source font. Our approach tackles these issues by leveraging recent advancements in Chinese character representation research to pre-train a robust content representation model. This model not only handles unseen characters but also generalizes to non-existent ones, a capability absent in traditional image-to-image translation. We further propose a Transformer-based Style Filter that not only accurately captures stylistic features from reference images but also handles any combination of them, fostering greater convenience for practical automated font generation applications. Additionally, we incorporate content loss with commonly used pixel- and perceptual-level losses to refine the generated results from a comprehensive perspective. Extensive experiments validate the effectiveness of our method, particularly its ability to handle unseen characters, demonstrating significant performance gains over existing state-of-the-art methods.

We present a computational approach for unfolding 3D shapes isometrically into the plane as a single patch without overlapping triangles. This is a hard, sometimes impossible, problem, which existing methods are forced to soften by allowing for map distortions or multiple patches. Instead, we propose a geometric relaxation of the problem: We modify the input shape until it admits an overlap-free unfolding. We achieve this by locally displacing vertices and collapsing edges, guided by the unfolding process. We validate our algorithm quantitatively and qualitatively on a large dataset of complex shapes and show its proficiency by fabricating real shapes from paper.

In this paper, we address the problem of plausible object placement for the challenging task of realistic image composition. We propose DiffPop, the first framework that utilizes plausibility-guided denoising diffusion probabilistic model to learn the scale and spatial relations among multiple objects and the corresponding scene image. First, we train an unguided diffusion model to directly learn the object placement parameters in a self-supervised manner. Then, we develop a human-in-the-loop pipeline which exploits human labeling on the diffusion-generated composite images to provide the weak supervision for training a structural plausibility classifier. The classifier is further used to guide the diffusion sampling process towards generating the plausible object placement. Experimental results verify the superiority of our method for producing plausible and diverse composite images on the new Cityscapes-OP dataset and the public OPA dataset, as well as demonstrate its potential in applications such as data augmentation and multi-object placement tasks. Our dataset and code will be released.

Recent advancements in generative models have enabled image editing very effective with impressive results. By extending this progress to 3D geometry models, we introduce iShapEditing, a novel framework for 3D shape editing which is applicable to both generated and real shapes. Users manipulate shapes by dragging handle points to corresponding targets, offering an intuitive and intelligent editing interface. Leveraging the Triplane Diffusion model and robust intermediate feature correspondence, our framework utilizes classifier guidance to adjust noise representations during sampling process, ensuring alignment with user expectations while preserving plausibility. For real shapes, we employ shape predictions at each time step alongside a DDPM-based inversion algorithm to derive their latent codes, facilitating seamless editing. iShapEditing provides effective and intelligent control over shapes without the need for additional model training or fine-tuning. Experimental examples demonstrate the effectiveness and superiority of our method in terms of editing accuracy and plausibility.

We introduce LGSur-Net, an end-to-end deep learning architecture, engineered for the upsampling of sparse point clouds. LGSur-Net harnesses a trainable Gaussian local representation by positioning a series of Gaussian functions on an oriented plane, complemented by the optimization of individual covariance matrices. The integration of parametric factors allows for the encoding of the plane's rotational dynamics and Gaussian weightings into a linear transformation matrix. Then we extract the feature maps from the point cloud and its adjoining edges and learn the local Gaussian depictions to accurately model the shape's local geometry through an attention-based network. The Gaussian representation's inherent high-order continuity endows LGSur-Net with the natural ability to predict surface normals and support upsampling to any specified resolution. Comprehensive experiments validate that LGSur-Net efficiently learns from sparse data inputs, surpassing the performance of existing state-of-the-art upsampling methods. Our code is publicly available at https://github.com/Rangiant5b72/LGSur-Net.