DeepJDOT: Deep Joint Distribution Optimal Transport for Unsupervised Domain Adaptation

Abstract

In computer vision, one is often confronted with problems of domain shifts, which occur when one applies a classifier trained on a source dataset to target data sharing similar characteristics (e.g. same classes), but also different latent data structures (e.g. different acquisition conditions). In such a situation, the model will perform poorly on the new data, since the classifier is specialized to recognize visual cues specific to the source domain. In this work we explore a solution, named DeepJDOT, to tackle this problem: through a measure of discrepancy on joint deep representations/labels based on optimal transport, we not only learn new data representations aligned between the source and target domain, but also simultaneously preserve the discriminative information used by the classifier. We applied DeepJDOT to a series of visual recognition tasks, where it compares favorably against state-of-the-art deep domain adaptation methods.

European Conference on Computer Vision 2018 (ECCV-2018)

References in corpus (7)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• VisDA: The Visual Domain Adaptation Challenge
• Unsupervised Image-to-Image Translation Networks
• Wasserstein Distance Guided Representation Learning for Domain Adaptation
• Stochastic Optimization for Large-scale Optimal Transport
• Return of Frustratingly Easy Domain Adaptation
• Large-Scale Optimal Transport and Mapping Estimation

Cited by in corpus (74)

• Artificial Intelligence in the Creative Industries: A Review
• A review of domain adaptation without target labels
• Domain Adaptive Ensemble Learning
• TimeMatch: Unsupervised Cross-Region Adaptation by Temporal Shift Estimation
• Dynamic Instance Domain Adaptation
• A Sentinel-2 multi-year, multi-country benchmark dataset for crop classification and segmentation with deep learning
• An Entropic Optimal Transport Loss for Learning Deep Neural Networks under Label Noise in Remote Sensing Images
• Unbalanced minibatch Optimal Transport; applications to Domain Adaptation
• Cross-domain Self-supervised Learning for Domain Adaptation with Few Source Labels
• An Unsupervised Domain Adaptation Model based on Dual-module Adversarial Training
• Rates of Estimation of Optimal Transport Maps using Plug-in Estimators via Barycentric Projections
• DAOT: Domain-Agnostically Aligned Optimal Transport for Domain-Adaptive Crowd Counting
• Unsupervised Domain Adaptation for LiDAR Panoptic Segmentation
• Unsupervised Neural Domain Adaptation for Document Image Binarization
• SALUDA: Surface-based Automotive Lidar Unsupervised Domain Adaptation
• Cross-hospital Sepsis Early Detection via Semi-supervised Optimal Transport with Self-paced Ensemble
• Learning with minibatch Wasserstein : asymptotic and gradient properties
• Wasserstein-2 Generative Networks
• Minibatch optimal transport distances; analysis and applications
• Prototypical Partial Optimal Transport for Universal Domain Adaptation
• Towards Unsupervised Domain Adaptation via Domain-Transformer
• SCLAiR : Supervised Contrastive Learning for User and Device Independent Airwriting Recognition
• DAPPER: Label-Free Performance Estimation after Personalization for Heterogeneous Mobile Sensing
• Online Adaptation to Label Distribution Shift
• Multi-Source Domain Adaptation through Dataset Dictionary Learning in Wasserstein Space
• Respecting Domain Relations: Hypothesis Invariance for Domain Generalization
• A Unified Joint Maximum Mean Discrepancy for Domain Adaptation
• On Scalable and Efficient Computation of Large Scale Optimal Transport
• Bridging the Domain Gap for Flight-Ready Spaceborne Vision
• Learning to Match Distributions for Domain Adaptation
• Learning a Domain-Invariant Embedding for Unsupervised Domain Adaptation Using Class-Conditioned Distribution Alignment
• Set Representation Learning with Generalized Sliced-Wasserstein Embeddings
• Unsupervised Domain Adaptation for Image Classification via Structure-Conditioned Adversarial Learning
• Learning Depth from Monocular Videos Using Synthetic Data: A Temporally-Consistent Domain Adaptation Approach
• Secure Domain Adaptation with Multiple Sources
• Reducing the Covariate Shift by Mirror Samples in Cross Domain Alignment
• G2DA: Geometry-Guided Dual-Alignment Learning for RGB-Infrared Person Re-Identification
• Flow-based Alignment Approaches for Probability Measures in Different Spaces
• Unsupervised Domain Adaptation with Progressive Domain Augmentation
• On Transportation of Mini-batches: A Hierarchical Approach
• Improving Mini-batch Optimal Transport via Partial Transportation
• Joint Semantic Domain Alignment and Target Classifier Learning for Unsupervised Domain Adaptation
• Federated Multi-Target Domain Adaptation
• Entropic Gromov-Wasserstein between Gaussian Distributions
• Cross-Language Aphasia Detection using Optimal Transport Domain Adaptation
• Domain Adaptation for Sentiment Analysis Using Increased Intraclass Separation
• Deep Optimal Transport for Domain Adaptation on SPD Manifolds
• Multiple Classifiers Based Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
• Dataset Dynamics via Gradient Flows in Probability Space
• Hidden Covariate Shift: A Minimal Assumption For Domain Adaptation
• SALT: Subspace Alignment as an Auxiliary Learning Task for Domain Adaptation
• Universal Domain Adaptation Benchmark for Time Series Data Representation
• Mining Label Distribution Drift in Unsupervised Domain Adaptation
• Large-Scale Optimal Transport via Adversarial Training with Cycle-Consistency
• Deep Inertial Navigation using Continuous Domain Adaptation and Optimal Transport
• Incremental Unsupervised Domain-Adversarial Training of Neural Networks
• Deep Unsupervised Domain Adaptation for Time Series Classification: a Benchmark
• On Label Shift in Domain Adaptation via Wasserstein Distance
• Matching Distributions via Optimal Transport for Semi-Supervised Learning
• Regularized Wasserstein Means for Aligning Distributional Data
• Convergence of Gaussian-smoothed optimal transport distance with sub-gamma distributions and dependent samples
• Unsupervised Latent Space Translation Network
• Conditional Bures Metric for Domain Adaptation
• Multi-Purposing Domain Adaptation Discriminators for Pseudo Labeling Confidence
• Towards Unsupervised Sketch-based Image Retrieval
• Deep Joint Distribution Optimal Transport for Universal Domain Adaptation on Time Series
• Geometry-Aware Unsupervised Domain Adaptation for Stereo Matching
• Explanatory Model Monitoring to Understand the Effects of Feature Shifts on Performance
• A Family of Pairwise Multi-Marginal Optimal Transports that Define a Generalized Metric
• Domain Adaptation and Active Learning for Fine-Grained Recognition in the Field of Biodiversity
• Unsupervised neural adaptation model based on optimal transport for spoken language identification
• Regularized Optimal Transport for Dynamic Semi-supervised Learning
• Graph Neural Networks for UnsupervisedDomain Adaptation of Histopathological ImageAnalytics
• Dynamic Flows on Curved Space Generated by Labeled Data