Unsupervised Accuracy Estimation of Deep Visual Models using Domain-Adaptive Adversarial Perturbation without Source Samples.
JoonHo LeeJae Oh WooHankyu MoonKwonho LeePublished in: ICCV (2023)
Keyphrases
- statistical models
- high accuracy
- parameter estimation
- domain models
- error analysis
- parametric models
- visual tasks
- neural network
- experimental data
- complex systems
- unsupervised learning
- data driven
- image classification
- semi supervised
- probabilistic model
- image sequences
- prediction accuracy
- computational cost
- computational models
- prior knowledge
- pairwise
- multi agent
- unseen data
- models built