RECORD
FDS: Feedback-guided Domain Synthesis with Multi-Source Conditional Diffusion Models for Domain Generalization
Computer Vision and Pattern Recognition · Submitted 4 Jul 2024
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Our comprehensive evaluations demonstrate that this methodology sets new benchmarks in domain generalization performance across a range of challenging datasets, effectively managing diverse types of domain shifts.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Domain Generalization techniques aim to enhance model robustness by simulating novel data distributions during training, typically through various augmentation or stylization strategies. However, these methods frequently suffer from limited control over the diversity of generated images and lack assurance that these images span distinct distributions. To address these challenges, we propose FDS, Feedback-guided Domain Synthesis, a novel strategy that employs diffusion models to synthesize novel, pseudo-domains by training a single model on all source domains and performing domain mixing based on learned features. By incorporating images that pose classification challenges to models trained on original samples, alongside the original dataset, we ensure the generation of a training set that spans a broad distribution spectrum. Our comprehensive evaluations demonstrate that this methodology sets new benchmarks in domain generalization performance across a range of challenging datasets, effectively managing diverse types of domain shifts. The code can be found at: \url{https://github.com/Mehrdad-Noori/FDS.git}.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Domain Generalization techniques aim to enhance model robustness by simulating novel data distributions during training, typically through various augmentation or stylization strategies. | background 1.00 | 0.03 | 0.02 | asserted 0.99 |
| s1 | However, these methods frequently suffer from limited control over the diversity of generated images and lack assurance that these images span distinct distributions. | gap 0.86 | 0.06 | 0.02 | asserted 0.90 |
| s2 | To address these challenges, we propose FDS, Feedback-guided Domain Synthesis, a novel strategy that employs diffusion models to synthesize novel, pseudo-domains by training a single model on all source domains and performing domain mixing based on learned features. | contribution 0.73 | 0.88 | 0.08 | asserted 0.99 |
| s3 | By incorporating images that pose classification challenges to models trained on original samples, alongside the original dataset, we ensure the generation of a training set that spans a broad distribution spectrum. | method 0.97 | 0.80 | 0.06 | asserted 0.99 |
| s4 | Our comprehensive evaluations demonstrate that this methodology sets new benchmarks in domain generalization performance across a range of challenging datasets, effectively managing diverse types of domain shifts. | result 0.99 | 0.95 | 0.88 | asserted 1.00 |
| s5 | The code can be found at: \url{https://github.com/Mehrdad-Noori/FDS.git}. | contribution 0.54 | 0.23 | 0.01 | asserted 0.88 |
What was examined, and what was not
- arXiv listing abstract v3
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Domain Generalization techniques aim to enhance model robustness by simulating novel data distributions during training, typically through various augmentation or stylization strategies. However, these methods frequently suffer from limited control over the diversity of generated images and lack assurance that these images span distinct distributions. To address these challenges, we propose FDS, Feedback-guided Domain Synthesis, a novel strategy that employs diffusion models to synthesize novel, pseudo-domains by training a single model on all source domains and performing domain mixing based on learned features. By incorporating images that pose classification challenges to models trained on original samples, alongside the original dataset, we ensure the generation of a training set that spans a broad distribution spectrum. Our comprehensive evaluations demonstrate that this methodology sets new benchmarks in domain generalization performance across a range of challenging datasets, effectively managing diverse types of domain shifts. The code can be found at: \url{https://github.com/Mehrdad-Noori/FDS.git}.
Human review
How the literature uses it
1 arXiv paper cite this in 1 sentence
Papers citing it
Cites it as background 1 paper
-
Generative-based Data Augmentation This technique is widely used to expand training datasets and prevent overfitting [ref] .
Generalize or Detect? Towards Robust Semantic Segmentation Under Multiple Distribution Shifts · Related Work · 0.97
What it cites · 4 sentences without a confident label
Compares with 1 paper
-
Furthermore, when comparing with previous SOTA methods, our method outperforms CDGA [ref] by 0.3% on PACS, 0.2% on VLCS, and 2.9% on OfficeHome.
Cross Domain Generative Augmentation: Domain Generalization with Latent Diffusion Models · Comparison with the State-of-the-art · 1.00
Cites as background 7 papers
-
In another study, CDGA [ref] uses diffusion models to generate synthetic images that fill the gap between different domain pairs through a simple interpolation.
Cross Domain Generative Augmentation: Domain Generalization with Latent Diffusion Models · Diffusion Models · 0.89
-
Innovations like denoising diffusion probabilistic models (DDPMs) [ref] and denoising diffusion implicit models (DDIMs) [ref] have significantly sped up the image generation process.
Denoising Diffusion Implicit Models · Diffusion Models · 0.87
-
Recently, various techniques have been developed to tackle this problem [ref] , including domain aligning [ref] , meta-learning [ref] , data augmentation [ref] , ensemble learning [ref] , self-supervised learning [ref] , and regularization methods [ref] .
Improving out-of-distribution generalization via multi-task self-supervised pretraining · Introduction · 0.96
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.