Cyc3D: Evaluating Cyclic Structural Stability and Asset Usability in Image-to-3D Generation
Computer Vision and Pattern Recognition
Summary
The authors created Cyc3D, a new benchmark to better evaluate how well computer models generate 3D objects from images. Unlike previous tests that only check if the generated views look right, Cyc3D also measures if the 3D object stays consistent and usable from different angles and if it has good quality for real graphics work. They introduced a way to repeatedly check the object by rendering it from new views and seeing if it changes or drifts in shape or meaning. When testing five popular systems, the authors found that while some perform better than others, all still struggle to make fully stable and reliable 3D objects. This shows there is still work to do before these models truly understand 3D objects well.
Authors
Liwen Zhang
Abstract
Image-conditioned 3D generation has advanced rapidly, yet existing evaluation protocols largely judge rendered-view plausibility and semantic alignment, overlooking whether a generator forms a stable 3D interpretation and produces assets usable in graphics pipelines. We introduce Cyc3D, a multidimensional benchmark that evaluates image-to-3D generation along two complementary axes: Cross-View Object Consistency and Representation Quality. At the asset level, Cyc3D measures whether object identity remains semantically coherent across rendered viewpoints. At the model level, we propose View-Cycle Structural Consistency, a closed-loop render-regenerate-align protocol that repeatedly re-observes a generated asset from novel views and quantifies geometric, perceptual, and semantic drift across generations. To assess native asset usability beyond rendered appearance, Cyc3D further evaluates geometric structure, reference-image fidelity, mesh discretization and efficiency, and UV parameterization quality. Together, these diagnostics expose failures obscured by a single perceptual score and provide interpretable evidence of both model instability and representation defects. Experiments on five representative image-to-3D systems show that closed-source feed-forward models consistently outperform open-source optimization-based baselines in geometric fidelity, mesh quality, and cycle stability. Nevertheless, even the strongest methods achieve cycle-stability scores below 48, revealing a persistent gap between visually plausible generation and robust 3D object understanding.