Dress Anyone : Automatic Physically-Based Garment Pattern Refitting

TL;DR

Dress Anyone addresses the challenge of automatically tailoring garments to diverse body shapes by jointly optimizing a physically simulated 3D drape and its 2D sewing patterns through differentiable simulation. The approach fits a target body, generates a corresponding 3D shape target, and refines the sewing patterns via gradient-based optimization, guided by a weighted loss that balances shape fidelity and design preservation. A Green-coordinate control cage and a multi-term loss enable stable, design-preserving refits, achieving higher pattern quality and broader body compatibility than prior work. The resulting simulation-ready 3D garments and manufacturable sewing patterns support downstream applications in games, virtual try-on, and real-world clothing fabrication, enabling scalable automatic tailoring across a wide range of body types, including fantastical shapes.

Abstract

Well-fitted clothing is essential for both real and virtual garments to enable self-expression and accurate representation for a large variety of body types. Common practice in the industry is to provide a pre-made selection of distinct garment sizes such as small, medium and large. While these may cater to certain groups of individuals that fall within this distribution, they often exclude large sections of the population. In contrast, individually tailored clothing offers a solution to obtain custom-fit garments that are tailored to each individual. However, manual tailoring is time-consuming and requires specialized knowledge, prohibiting the approach from being applied to produce fitted clothing at scale. To address this challenge, we propose a novel method leveraging differentiable simulation for refitting and draping 3D garments and their corresponding 2D pattern panels onto a new body shape, enabling a workflow where garments only need to be designed once, in a single size, and they can be automatically refitted to support numerous body size and shape variations. Our method enables downstream applications, where our optimized 3D drape can be directly ingested into game engines or other applications. Our 2D sewing patterns allow for accurate physics-based simulations and enables manufacturing clothing for the real world.

Figures (7)

• Figure 1: Given a draped input garment and body, Dress Anyone produces refitted 2D patterns and a 3D draped garment to fit a provided target body shape. We first fit a template body model to the target body to make the method robust against topology differences between body meshes. We the produce a target 3D garment shape fernando which we use as a target shape. We then use differentiable simulation to optimize for a refitted 2D pattern that fits the target body. We leverage a robust and efficient control cage formulation which preserves the garment design. Our refitted garments can be used for several downstream applications such as novel motion generation.
• Figure 2: Animated sequences of refitted clothing. Our method produces custom fit garments, allowing them to be directly draped and simulated. The incorporation of 2D sewing patterns provides an accurate rest shape which significantly improves garment simulation realism. 3D draped geometries do not provide an accurate rest shape because they frequently contain deformations such as stretching or sagging under gravity which are already integrated into the mesh. This can lead to inaccuracies during the simulation process. Additionally, it allows for our garments to be manufactured from real fabrics. We showcase a select number of frames of a yoga sequence, demonstrating that the fitted garments allow for rich dynamics and diversity in body pose.
• Figure 3: 3D Ablation. We provide a visual ablation of the resulting 3D physically simulated drapes when omitting loss term regularizers. Far left shows a visualization of the individual garment panels, highlighting the seam locations used for the seam location matching. We then show the target 3D shape obtained by fernando which does not produce 2D sewing patterns. Our full method closely matches the target drape, including the faithful recreation of the ankle cuffs, with custom fit sewing patterns. Note that we do not expect an exact match since the target drape is not physically simulated under external forces such as gravity or body and cloth self collisions. Omitting either of the regularizers produces lower quality fits that do not match the target as closely. Note especially that the ankle cuffs are only preserved with our full approach. The target location of the crotch seam line (highlighted by the red line) is only matched with our full method.
• Figure 4: 2D Ablation. We assess the importance of our proposed symmetry enforcing approach (ski suit, top) and boundary curvature regularization (pants, bottom) by analyzing the sewing patterns in 2D space. Top : Symmetry enforcement naturally leads to realistic garments patterns to how an experienced tailor would produce them. Bottom : We additionally see the importance of boundary curvature regularization. Omitting this term produces curved pattern boundaries, drastically changing the intended design of the original fit.
• Figure 5: We demonstrate a real world example of a garment refitted using our approach where we manufactured a physical shirt given the refitted sewing patterns. Left shows the original garment fitted to the original body. Middle shows our virtual refitted garment which is physically draped on the new body. The body geometry was obtained using a 3D scan. Right shows the real garment draped on the real person from whom the scan was taken. Note how the real drape matches the virtual resized garment.