Humans and traditional computer vision methods rely on a diverse set of monocular cues to infer 3D structure from a single image, such as shading, texture, silhouette, etc. While recent deep generative models have dramatically advanced single-image 3D generation, it remains unclear which image cues these methods actually exploit. We introduce Cue3D, the first comprehensive, model-agnostic framework for quantifying the influence of individual image cues in single-image 3D generation. Our unified benchmark evaluates seven state-of-the-art methods, spanning regression-based, multi-view, and native 3D generative paradigms.

Reference-driven image completion, which restores missing regions in a target view using additional images, is particularly challenging when the target view differs significantly from the references. Existing generative methods rely solely on diffusion priors and, without geometric cues such as camera pose or depth, often produce misaligned or implausible content. We propose GeoComplete, a novel framework that incorporates explicit 3D structural guidance to enforce geometric consistency in the completed regions, setting it apart from prior image-only approaches. GeoComplete introduces two key ideas: conditioning the diffusion process on projected point clouds to infuse geometric information, and applying target-aware masking to guide the model toward relevant reference cues. The framework features a dual-branch diffusion architecture.

In this section, we first present an overview of 4 different architectures for neural implicit scene representations anddetails ofMulti-Res. See Figure 1 for an overview over the architectures. More specifically, each grid contains up toT feature vectors with dimensionalityF. We further reportNormal Consistencyfor the Replica dataset following [9,13,18,19,23,32] as near-perfect ground truth is available. We observe that using more input views for training improves reconstruction quality.

Grids in Section 1.1 and provide details of the depth loss In the following, we provide details for Multi-Res. For our single MLP architecture, we use an 8-layer MLP with hidden dimension 256. We use a two-layer MLP with hidden dimension 256 for the SDF prediction for both, Single-Res. For the DTU dataset [1], we follow the official evaluation protocol and report the reconstruction quality with: Accuracy, Completeness and Chamfer Distance . Distance is the mean of Accuracy and Completeness .

We address the task of open-world class-agnostic object detection, i.e., detecting every object in an image by learning from a limited number of base object classes. State-of-the-art RGB-based models suffer from overfitting the training classes and often fail at detecting novel-looking objects. This is because RGB-based models primarily rely on appearance similarity to detect novel objects and are also prone to overfitting short-cut cues such as textures and discriminative parts. To address these shortcomings of RGB-based object detectors, we propose incorporating geometric cues such as depth and normals, predicted by general-purpose monocular estimators. Specifically, we use the geometric cues to train an object proposal network for pseudo-labeling unannotated novel objects in the training set. Our resulting Geometry-guided Open-world Object Detector (GOOD) significantly improves detection recall for novel object categories and already performs well with only a few training classes. Using a single "person" class for training on the COCO dataset, GOOD surpasses SOTA methods by 5.0% AR@100, a relative improvement of 24%.