Designing and manipulating virtual human heads is essential across various applications, including AR, VR, gaming, human-computer interaction and VFX. Traditional graphic-based approaches require manual effort and resources to achieve accurate representation of human heads. While modern deep learning techniques can generate and edit highly photorealistic images of faces, their focus remains predominantly on 2D facial images. This limitation makes them less suitable for 3D applications. Recognizing the vital role of editing within the UV texture space as a key component in the 3D graphics pipeline, our work focuses on this aspect to benefit graphic designers by providing enhanced control and precision in appearance manipulation. Research on existing methods within the UV texture space is limited, complex, and poses challenges. In this paper, we introduce SemUV: a simple and effective approach using the FFHQ-UV dataset for semantic manipulation directly within the UV texture space. We train a StyleGAN model on the publicly available FFHQ-UV dataset, and subsequently train a boundary for interpolation and semantic feature manipulation. Through experiments comparing our method with 2D manipulation technique, we demonstrate its superior ability to preserve identity while effectively modifying semantic features such as age, gender, and facial hair. Our approach is simple, agnostic to other 3D components such as structure, lighting, and rendering, and also enables seamless integration into standard 3D graphics pipelines without demanding extensive domain expertise, time, or resources.

This paper strives for image editing via generative models. Flow Matching is an emerging generative modeling technique that offers the advantage of simple and efficient training. Simultaneously, a new transformer-based U-ViT has recently been proposed to replace the commonly used UNet for better scalability and performance in generative modeling. Hence, Flow Matching with a transformer backbone offers the potential for scalable and high-quality generative modeling, but their latent structure and editing ability are as of yet unknown. Hence, we adopt this setting and explore how to edit images through latent space manipulation. We introduce an editing space, which we call $u$-space, that can be manipulated in a controllable, accumulative, and composable manner. Additionally, we propose a tailored sampling solution to enable sampling with the more efficient adaptive step-size ODE solvers. Lastly, we put forth a straightforward yet powerful method for achieving fine-grained and nuanced editing using text prompts. Our framework is simple and efficient, all while being highly effective at editing images while preserving the essence of the original content. Our code will be publicly available at https://taohu.me/lfm/

While natural language offers a convenient shared interface for humans and robots, enabling robots to interpret and follow language commands remains a longstanding challenge in manipulation. A crucial step to realizing a performant instruction-following robot is achieving semantic manipulation, where a robot interprets language at different specificities, from high-level instructions like "Pick up the stuffed animal" to more detailed inputs like "Grab the left ear of the elephant." To tackle this, we propose Keypoints + Instructions to Execution (KITE), a two-step framework for semantic manipulation which attends to both scene semantics (distinguishing between different objects in a visual scene) and object semantics (precisely localizing different parts within an object instance). KITE first grounds an input instruction in a visual scene through 2D image keypoints, providing a highly accurate object-centric bias for downstream action inference. Provided an RGB-D scene observation, KITE then executes a learned keypoint-conditioned skill to carry out the instruction. The combined precision of keypoints and parameterized skills enables fine-grained manipulation with generalization to scene and object variations. Empirically, we demonstrate KITE in 3 real-world environments: long-horizon 6-DoF tabletop manipulation, semantic grasping, and a high-precision coffee-making task. In these settings, KITE achieves a 75%, 70%, and 71% overall success rate for instruction-following, respectively. KITE outperforms frameworks that opt for pre-trained visual language models over keypoint-based grounding, or omit skills in favor of end-to-end visuomotor control, all while being trained from fewer or comparable amounts of demonstrations. Supplementary material, datasets, code, and videos can be found on our website: http://tinyurl.com/kite-site.