Search for: All records

We propose the Large View Synthesis Model (LVSM), a novel transformer-based approach for scalable and generalizable novel view synthesis from sparse-view inputs. We introduce two architectures: (1) an encoder-decoder LVSM, which encodes input image tokens into a fixed number of 1D latent tokens, functioning as a fully learned scene representation, and decodes novel-view images from them; and (2) a decoder-only LVSM, which directly maps input images to novel-view outputs, completely eliminating intermediate scene representations. Both models bypass the 3D inductive biases used in previous methods—from 3D representations (e.g., NeRF, 3DGS) to network designs (e.g., epipolar projections, plane sweeps)—addressing novel view synthesis with a fully data-driven approach. While the encoder-decoder model offers faster inference due to its independent latent representation, the decoder-only LVSM achieves superior quality, scalability, and zero-shot generalization, outperforming previous state-of-the-art methods by 1.5 to 3.5 dB PSNR. Comprehensive evaluations across multiple datasets demonstrate that both LVSM variants achieve state-of-the-art novel view synthesis quality. Notably, our models surpass all previous methods even with reduced computational resources (1-2 GPUs). 

Lithospheric shortening can be described by one of two end-member modes: indentation of the lithosphere and subduction of the lithospheric mantle. Deciphering the difference between these modes is crucial in the interpretation of past and present orogens and in predicting their structural architecture at depth. It is therefore important to establish how observable upper crustal proxies reflect deep lithospheric kinematics and dynamics. Over the last few decades, geological and geophysical data have provided valuable constraints on the northern margin of the Tibetan Plateau. This margin is defined by the Qilian Shan thrust belt, which developed in response to the far-field convergence between the Indian and Eurasian plates. The primary mechanism for this development is the southward subduction of the Asian lithospheric mantle beneath the Tibetan Plateau. We conducted numerical modelling to simulate the kinematics and response of the upper crust to the southward subduction of the lithospheric mantle. Our results show that subduction of the lithospheric mantle can result in upper crustal deformation that matches the records in the Qilian Shan, where pure shear shortening alone does not generate similar upper crust proxies, including the broad width and architecture of the bivergent orogenic wedge, the timing of fault initiation and evolution, seismicity and fault activity, the topography and geomorphology. The geometry of the subducting lithosphere impacts the width and asymmetry of the bivergent orogenic wedge. Our results demonstrate how records of crustal strain can be used to better interpret the deep structural architecture of past and present orogenic wedges.