SurfaceNet: Adversarial SVBRDF Estimation from a Single Image
arXiv:2107.11298 · doi:10.1109/ICCV48922.2021
Abstract
In this paper we present SurfaceNet, an approach for estimating spatially-varying bidirectional reflectance distribution function (SVBRDF) material properties from a single image. We pose the problem as an image translation task and propose a novel patch-based generative adversarial network (GAN) that is able to produce high-quality, high-resolution surface reflectance maps. The employment of the GAN paradigm has a twofold objective: 1) allowing the model to recover finer details than standard translation models; 2) reducing the domain shift between synthetic and real data distributions in an unsupervised way. An extensive evaluation, carried out on a public benchmark of synthetic and real images under different illumination conditions, shows that SurfaceNet largely outperforms existing SVBRDF reconstruction methods, both quantitatively and qualitatively. Furthermore, SurfaceNet exhibits a remarkable ability in generating high-quality maps from real samples without any supervision at training time.
References in corpus (2)
Cited by in corpus (35)
- Vision-Language Models for Medical Report Generation and Visual Question Answering: A Review
- Comparison of fine-tuning strategies for transfer learning in medical image classification
- Using generative AI to investigate medical imagery models and datasets
- AutoFuse: Automatic Fusion Networks for Deformable Medical Image Registration
- Pixel-Level Clustering Network for Unsupervised Image Segmentation
- Detection, Instance Segmentation, and Classification for Astronomical Surveys with Deep Learning (DeepDISC): Detectron2 Implementation and Demonstration with Hyper Suprime-Cam Data
- Representing Long Volumetric Video with Temporal Gaussian Hierarchy
- Digitizing Historical Balance Sheet Data: A Practitioner's Guide
- Topological Data Analysis in smart manufacturing: State of the art and futuredirections
- A review of advancements in low-light image enhancement using deep learning
- Significantly improving zero-shot X-ray pathology classification via fine-tuning pre-trained image-text encoders
- Spatiotemporal information conversion machine for time-series prediction
- Deep Image Compression Using Scene Text Quality Assessment
- Consistent Attack: Universal Adversarial Perturbation on Embodied Vision Navigation
- SHREC 2022: Fitting and recognition of simple geometric primitives on point clouds
- Impact of color and mixing proportion of synthetic point clouds on semantic segmentation
- Mpox-AISM: AI-Mediated Super Monitoring for Mpox and Like-Mpox
- Glioma subtype classification from histopathological images using in-domain and out-of-domain transfer learning: An experimental study
- ESTformer: Transformer utilising spatiotemporal dependencies for electroencephalogram super-resolution
- Towards Universal Texture Synthesis by Combining Texton Broadcasting with Noise Injection in StyleGAN-2
- MARF: The Medial Atom Ray Field Object Representation
- MLLM-Based UI2Code Automation Guided by UI Layout Information
- FarNet-II: An improved solar far-side active region detection method
- Real-Time Idling Vehicles Detection using Combined Audio-Visual Deep Learning
- Towards a vision foundation model for comprehensive assessment of Cardiac MRI
- Classifier Chain Networks for Multi-Label Classification
- Towards Developing Socially Compliant Automated Vehicles: Advances, Expert Insights, and A Conceptual Framework
- Cross-Sign Language Transfer Learning Using Domain Adaptation with Multi-scale Temporal Alignment
- FA-Seg: A Fast and Accurate Diffusion-Based Method for Open-Vocabulary Segmentation
- A Deep Learning Pipeline for Solid Waste Detection in Remote Sensing Images
- Metamorphic Testing of Multimodal Human Trajectory Prediction
- Multi-modal Traffic Scenario Generation for Autonomous Driving System Testing
- Text-based Animatable 3D Avatars with Morphable Model Alignment
- Improved Single Camera BEV Perception Using Multi-Camera Training
- NeMo: A Neuron-Level Modularizing-While-Training Approach for Decomposing DNN Models