Adversarial Semantic Hallucination for Domain Generalized Semantic Segmentation

  title={Adversarial Semantic Hallucination for Domain Generalized Semantic Segmentation},
  author={Gabriel Tjio and Ping Liu and Joey Tianyi Zhou and Rick Siow Mong Goh},
  journal={2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  • Gabriel TjioPing Liu R. Goh
  • Published 8 June 2021
  • Computer Science
  • 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
Convolutional neural networks typically perform poorly when the test (target domain) and training (source domain) data have significantly different distributions. While this problem can be mitigated by using the target domain data to align the source and target domain feature representations, the target domain data may be unavailable due to privacy concerns. Consequently, there is a need for methods that generalize well despite restricted access to target domain data during training. In this… 

Figures and Tables from this paper

Embedding Hallucination for Few-shot Language Fine-tuning

An Embedding Hallucination (EmbedHalluc) method is proposed, which generates auxiliary embedding-label pairs to expand the fine-tuning dataset, and shows that it outperforms other methods that address this over-fitting problem, such as common data augmentation, semi-supervised pseudo-labeling, and regularization.

Omni-supervised Facial Expression Recognition via Distilled Data

This paper proposes to perform omni-supervised learning by exploiting not only the labeled samples, but also the unlabeled data, which is relatively easy to obtain and can introduce more variations to improve the generality of the trained models.



Learning to Adapt Structured Output Space for Semantic Segmentation

A multi-level adversarial network is constructed to effectively perform output space domain adaptation at different feature levels and it is shown that the proposed method performs favorably against the state-of-the-art methods in terms of accuracy and visual quality.

Category-Level Adversarial Adaptation for Semantic Segmentation using Purified Features.

The designed network, CLAN, can not only disentangle the noisy factors and suppress their influences for target tasks but also utilize those purified features to conduct a more delicate level domain calibration simultaneously, i.e., global marginal distribution and local joint distribution alignment simultaneously.

Significance-Aware Information Bottleneck for Domain Adaptive Semantic Segmentation

This work equips the adversarial network with a "significance-aware information bottleneck (SIB)", to address the above problem, and shows that the proposed method can yield leading results compared with other feature-space alternatives.

ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation

This work proposes two novel, complementary methods using (i) entropy loss and (ii) adversarial loss respectively for unsupervised domain adaptation in semantic segmentation with losses based on the entropy of the pixel-wise predictions.

Classes Matter: A Fine-grained Adversarial Approach to Cross-domain Semantic Segmentation

This work proposes a fine-grained adversarial learning strategy for class-level feature alignment while preserving the internal structure of semantics across domains, which outperforms other global feature alignment based and class-wise alignment based counterparts.

Label-Driven Reconstruction for Domain Adaptation in Semantic Segmentation

This work presents an innovative framework, designed to mitigate the image translation bias and align cross-domain features with the same category by performing the target-to-source translation and reconstructing both source and target images from their predicted labels.

CyCADA: Cycle-Consistent Adversarial Domain Adaptation

A novel discriminatively-trained Cycle-Consistent Adversarial Domain Adaptation model that adapts representations at both the pixel-level and feature-level, enforces cycle-consistency while leveraging a task loss, and does not require aligned pairs is proposed.

Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation

A category-level adversarial network is introduced, aiming to enforce local semantic consistency during the trend of global alignment, to take a close look at the category- level data distribution and align each class with an adaptive adversarial loss.

Adversarial Style Mining for One-Shot Unsupervised Domain Adaptation

A novel Adversarial Style Mining approach, which combines the style transfer module and task-specific module into an adversarial manner, which achieves state-of-the-art adaptation performance under the challenging one-shot setting.

Unsupervised Domain Adaptation for Semantic Segmentation via Class-Balanced Self-training

This paper proposes a novel UDA framework based on an iterative self-training (ST) procedure, where the problem is formulated as latent variable loss minimization, and can be solved by alternatively generating pseudo labels on target data and re-training the model with these labels.