LAUSR

Existing convolutional neural networks have achieved great success in recovering SVBRDF maps from a single image. However, they mainly focus on handling low-resolution (e.g., 256 × 256) inputs. Ultra-high resolution (UHR) material maps are notoriously difficult to acquire by existing networks because: 1) finite computational resources set bounds for input receptive fields and output resolutions; 2) convolutional layers operate locally and lack the ability to capture long-range structural dependencies in UHR images. We propose an implicit neural reflectance model and a divide-and-conquer solution to address these two challenges simultaneously. We first crop an UHR image into low-resolution patches, each of which are processed by a local feature extractor (LFE) to extract important details. To fully exploit long-range spatial dependency and ensure global coherency, we incorporate a global feature extractor (GFE) and several coordinate-aware feature assembly (CAFA) modules into our pipeline. The GFE contains several lightweight material vision transformers that have a global receptive field at each scale and have the ability to infer long-term relationships in the material. After decoding globally coherent feature maps assembled by CAFA modules, the proposed end-to-end method is able to generate UHR SVBRDF maps from a single image with fine spatial details and consistent global structures.