SPIGAN: Privileged Adversarial Learning from Simulation
arXiv:1810.03756
Abstract
Deep Learning for Computer Vision depends mainly on the source of supervision.Photo-realistic simulators can generate large-scale automatically labeled syntheticdata, but introduce a domain gap negatively impacting performance. We propose anew unsupervised domain adaptation algorithm, called SPIGAN, relying on Sim-ulator Privileged Information (PI) and Generative Adversarial Networks (GAN).We use internal data from the simulator as PI during the training of a target tasknetwork. We experimentally evaluate our approach on semantic segmentation. Wetrain the networks on real-world Cityscapes and Vistas datasets, using only unla-beled real-world images and synthetic labeled data with z-buffer (depth) PI fromthe SYNTHIA dataset. Our method improves over no adaptation and state-of-the-art unsupervised domain adaptation techniques.
Accepted by ICLR 2019
References in corpus (17)
- Adam: A Method for Stochastic Optimization
- Distilling the Knowledge in a Neural Network
- Learning Transferable Features with Deep Adaptation Networks
- Deep Domain Confusion: Maximizing for Domain Invariance
- Learning to Discover Cross-Domain Relations with Generative Adversarial Networks
- FCNs in the Wild: Pixel-level Adversarial and Constraint-based Adaptation
- Domain Separation Networks
- Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes
- Unsupervised Image-to-Image Translation Networks
- Unsupervised Cross-Domain Image Generation
- Learning Domain-Invariant Subspace using Domain Features and Independence Maximization
- Unifying distillation and privileged information
- Learning Using Privileged Information: SVM+ and Weighted SVM
- Least Squares Generative Adversarial Networks
- Efficient Learning of Domain-invariant Image Representations
- Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
- Semantic Pose using Deep Networks Trained on Synthetic RGB-D