Hyden: A Hybrid Dual-Path Encoder for Monocular Geometry of High-resolution Images

We present a hybrid dual-path vision encoder (Hyden) for high-resolution monocular depth, point map and surface normal estimation, surpassing state-of-the-art accuracy with a fraction of the inference cost. The architecture pairs a lowresolution Vision Transformer branch for global context with a full-resolution CNN branch for fine details, fusing features via a lightweight MLP before decoding. By exploiting the linear scaling of CNNs and constraining transformer computation to a fixed resolution, the model delivers fast inference even on multimegapixel inputs. To overcome the scarcity of high-quality high-resolution supervision, we introduce a self-distillation framework that generates pseudo-labels from existing models at both lower resolution full images and high-resolution crops-global labels preserve geometric accuracy, while local labels capture sharper details. To demonstrate the flexibility of our approach, we integrate Hyden and our self-distillation method into DepthAnything-v2 for depth estimation and MoGe2 for surface normal and metric point map prediction, achieving stateof-the-art results on high-resolution benchmarks with the lowest inference latency among competing methods. 2 RELATED WORK 2.1 ZERO-SHOT MONOCULAR GEOMETRY ESTIMATION Traditional monocular models Bhat et al. (2021); Eigen et al. (2014); Li et al. (2022); Eigen & Fergus (2015); Saxena et al. ( 2008 ) were trained on single datasets for specific domains (e.g., indoor or street-view) and generalized poorly due to limited diversity and fixed camera setups.