RASP: Revisiting 3D Anamorphic Art for Shadow-Guided Packing of Irregular Objects

TL;DR

RASP addresses arranging arbitrarily shaped 3D objects inside a bounded volume using shadows from multiple viewpoints as supervision. It introduces a differentiable rendering pipeline with quaternion-based pose optimization and an $SDF$-based loss to enforce non-overlap and containment, combining $L_{sil}$, $L_{is}$, and $L_{ext}$ as $L_{total} = L_{sil} + L_{is} + \lambda L_{ext}$ with $\lambda=0.001$. The framework supports both packing and 3D part assembly, and extends to multi-view anamorphic art by adding texture optimization. Across experiments on irregular object datasets and the Fantastic Breaks collection, RASP achieves competitive occupancy and demonstrates versatile applications from concrete packing to artistic 3D illustrations, all without explicit 3D ground-truth supervision. This work opens new avenues for shadow-guided 3D understanding and design in applications spanning logistics, CAD, and creative visualization.

Abstract

Recent advancements in learning-based methods have opened new avenues for exploring and interpreting art forms, such as shadow art, origami, and sketch art, through computational models. One notable visual art form is 3D Anamorphic Art in which an ensemble of arbitrarily shaped 3D objects creates a realistic and meaningful expression when observed from a particular viewpoint and loses its coherence over the other viewpoints. In this work, we build on insights from 3D Anamorphic Art to perform 3D object arrangement. We introduce RASP, a differentiable-rendering-based framework to arrange arbitrarily shaped 3D objects within a bounded volume via shadow (or silhouette)-guided optimization with an aim of minimal inter-object spacing and near-maximal occupancy. Furthermore, we propose a novel SDF-based formulation to handle inter-object intersection and container extrusion. We demonstrate that RASP can be extended to part assembly alongside object packing considering 3D objects to be "parts" of another 3D object. Finally, we present artistic illustrations of multi-view anamorphic art, achieving meaningful expressions from multiple viewpoints within a single ensemble.

Figures (12)

• Figure 1: (a) An ensemble of arbitrary shapes casting the shadow of alphabets CVPR and numbers 2025, (b) a set of irregular objects packed inside a face-shaped container, (c) an assembly of parts of a vessel obtained through multi-view shadow guidance via RASP.
• Figure 2: Illustration of 3D Anamorphic Art (a, b): Portrait of Nikola Tesla and Bedřich Smetana by Patrick Proško and Shadow Art (c, d): Dirty White Trash and Wild Mood Swing by Tim Nobel and Sue Webster.
• Figure 3: An example depicting the packed state of a cuboidal container and the associated silhouettes/shadows.
• Figure 4: The proposed differentiable rendering-based pipeline for RASP. We use Adam optimizer with a learning rate of $1e-2$ to $1e-4$ over $1000$ iterations. Packing the Kitchen dataset ($106$ objects, $80000$ grid points) takes nearly $17$ minutes, and reassembly for each object in the Fantastic Breaks dataset ($2$ parts, $80000$ grid points) takes under $3$ minutes.
• Figure 5: (a) Target shadow a duck-shaped container, optimized arrangement with mesh intersections. Mesh enclosed within (b) sphere and (c) cube where the intersection region doesn’t include object meshes. (d) SDF of two non-intersecting and intersecting objects (intersecting points are marked in red), and SDF transformation within the container.