Learning Texture Invariant Representation for Domain Adaptation of Semantic Segmentation

By Myeongjin Kim, Hyeran Byun. Computer Vision and Pattern Recognition, 2020

309 citations recorded by the source corpus.

Since annotating pixel-level labels for semantic segmentation is laborious, leveraging synthetic data is an attractive solution. However, due to the domain gap between synthetic domain and real domain, it is challenging for a model trained with synthetic data to generalize to real data. In this paper, considering the fundamental difference between the two domains as the texture, we propose a method to adapt to the target domain's texture. First, we diversity the texture of synthetic images using a style transfer algorithm. The various textures of generated images prevent a segmentation model from overfitting to one specific (synthetic) texture. Then, we fine-tune the model with self-training to get direct supervision of the target texture. Our results achieve state-of-the-art performance and we analyze the properties of the model trained on the stylized dataset with extensive experiments.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/0a1077cee0fa709b2a0087cc25f8bc46884be159

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

ACDC: The Adverse Conditions Dataset with Correspondences for Semantic Driving Scene Understanding (2021, methodology): “A multitude of recent works [7, 15, 22, 23, 26, 41, 43, 44, 46, 48, 51, 53, 59, 60, 62, 65, 66] have focused on unsupervised domain adaptation (UDA) for semantic segmentation, but most of them are validated only on an artiﬁcial synthetic-to-real setting, using GTA5 [34] and SYNTHIA [36] as source…”
HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation (2022, unknown): “Several works also combine self-training and adversarial training [34, 37, 64, 86], minimize the entropy [7, 62, 63], refine boundaries [39], use a curriculum [12, 13, 82], or learn auxiliary tasks [8, 29, 63, 66].”
Exploring Robustness of Unsupervised Domain Adaptation in Semantic Segmentation (2021, background): “The rationale behind this setting is that i) recent state-of-the-art UDA methods in semantic segmentation [42, 17, 48, 44, 31] share similar spirits with BDL, so conducting pilot studies on this method would be representative; ii) a black-box attack assumes that the attacker can only access very lim”
DACS: Domain Adaptation via Cross-domain Mixed Sampling (2020, background): “Other approaches to aligning the source and target domains include channel-wise alignment of feature maps [41], done differently for ”stuff” classes and ”things” classes in [39], transfer of texture from target to source [18], and transfer of the low-frequency part of the spectrum in Fourier space f”
DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (2021, methodology): “Several methods also combine adversarial and self-training [37,40,74], train with auxiliary tasks [32, 73, 75], or perform test-time UDA [76].”
Give Me Your Trained Model: Domain Adaptive Semantic Segmentation without Source Data (2021, background): “For adversarial learning, numerous works focus on reducing the distribution misalignment in the image level [12, 13, 10, 14, 15, 5, 16], feature level [17, 8, 9], and output level [2–4 , 18].”
Where and How to Transfer: Knowledge Aggregation-Induced Transferability Perception for Unsupervised Domain Adaptation (2021, methodology): “…[33] 3 : 73 (cid:4) 10 (cid:6) 3 5 : 96 (cid:4) 10 (cid:6) 3 Ours versus PIT [32] 2 : 45 (cid:4) 10 (cid:6) 3 8 : 14 (cid:4) 10 (cid:6) 3 Ours versus LTIR [23] 6 : 58 (cid:4) 10 (cid:6) 3 8 : 14 (cid:4) 10 (cid:6) 3 Ours versus ELS-DA [12] 3 : 62 (cid:4) 10 (cid:6) 3 4 : 75 (cid:4) 10 (cid:6) 3”
PixMatch: Unsupervised Domain Adaptation via Pixelwise Consistency Training (2021, methodology): “used AvatarNet [45] rather than CycleGAN as in [33] or CycleGAN + StyleSwap as in [29].”
MetaCorrection: Domain-aware Meta Loss Correction for Unsupervised Domain Adaptation in Semantic Segmentation (2021, unknown): “Compared with domain alignment methods [40, 41, 20], the proposed method shows superior performance.”
A Review of Single-Source Deep Unsupervised Visual Domain Adaptation (2020, methodology): “[203, 204] proposed to use style transfer to randomize the input domains for better generalization performance.”