Filters








4,758 Hits in 4.1 sec

Reducing Domain Gap by Reducing Style Bias [article]

Hyeonseob Nam, HyunJae Lee, Jongchan Park, Wonjun Yoon, Donggeun Yoo
2021 arXiv   pre-print
Inspired by this, we propose to reduce the intrinsic style bias of CNNs to close the gap between domains.  ...  Our Style-Agnostic Networks (SagNets) disentangle style encodings from class categories to prevent style biased predictions and focus more on the contents.  ...  Our method is orthogonal to the majority of existing domain adaptation and generalization techniques that utilize Figure 1 : Our Style-Agnostic Network (SagNet) reduces style bias to reduce domain gap  ... 
arXiv:1910.11645v4 fatcat:ugujqj3jxrffjicyz5oouv5gna

Learning a Domain-Agnostic Visual Representation for Autonomous Driving via Contrastive Loss [article]

Dongseok Shim, H. Jin Kim
2021 arXiv   pre-print
DACL leads the neural network to learn domain-agnostic representation to overcome performance degradation when there exists a difference between training and test data distribution.  ...  In this paper, we propose a Domain-Agnostic Contrastive Learning (DACL) which is a two-stage unsupervised domain adaptation framework with cyclic adversarial training and contrastive loss.  ...  Stage 1: Bidirectional Style Transfer First, we train two style transfer networks G S→T and G T →S to bridge the gap between the source domain S and the target domain T in a bidirectional flow.  ... 
arXiv:2103.05902v1 fatcat:qfyess3ennb2hl436set44vrhe

Camera-aware Style Separation and Contrastive Learning for Unsupervised Person Re-identification [article]

Xue Li, Tengfei Liang, Yi Jin, Tao Wang, Yidong Li
2021 arXiv   pre-print
It can explicitly divide the learnable feature into camera-specific and camera-agnostic parts, reducing the influence of different cameras.  ...  However, most methods ignore the intra-class gap caused by camera style variance, and some methods are relatively complex and indirect although they try to solve the negative impact of the camera style  ...  As for the UDA methods, some of them use GAN [14, 15] for style transfer to narrow the gap between the source domain and target domain.  ... 
arXiv:2112.10089v1 fatcat:ygqc2volujfvbovsjbmqkycc4u

Style Variable and Irrelevant Learning for Generalizable Person Re-identification [article]

Haobo Chen, Chuyang Zhao, Kai Tu, Junru Chen, Yadong Li, Boxun Li
2022 arXiv   pre-print
Specifically, we design a Style Jitter Module (SJM) in SVIL. The SJM module can enrich the style diversity of the specific source domain and reduce the style differences of various source domains.  ...  In this paper, we first verify through an experiment that style factors are a vital part of domain bias.  ...  After the Global Average Pooling (GAP), we design K domain-specific classifiers ϕ k (•) and a domain agnostic classifier ϕ g (•) to learn the feature representations.  ... 
arXiv:2209.05235v1 fatcat:ji4wcdjgvvh7dapl44w7tdggjq

StyleMeUp: Towards Style-Agnostic Sketch-Based Image Retrieval [article]

Aneeshan Sain, Ayan Kumar Bhunia, Yongxin Yang, Tao Xiang, Yi-Zhe Song
2021 arXiv   pre-print
model truly style-agnostic.  ...  An effective SBIR model needs to explicitly account for this style diversity, crucially, to generalise to unseen user styles. To this end, a novel style-agnostic SBIR model is proposed.  ...  Research has flourished in recent years, where the main focus has been on addressing the sketch-photo domain gap [33, 16, 44] and data scarcity [5, 3, 10, 35, 12] .  ... 
arXiv:2103.15706v2 fatcat:ukbeu2bpujb53j3pzd5zbdldai

Robust Multi-Task Learning and Online Refinement for Spacecraft Pose Estimation across Domain Gap [article]

Tae Ha Park, Simone D'Amico
2022 arXiv   pre-print
This work presents Spacecraft Pose Network v2 (SPNv2), a Convolutional Neural Network (CNN) for pose estimation of noncooperative spacecraft across domain gap.  ...  This work also introduces Online Domain Refinement (ODR) which refines the parameters of the normalization layers of SPNv2 on the target domain images online at deployment.  ...  Acknowledgement This work is supported by the Air Force Office of Scientific Research (AFOSR) via Centauri under the project titled Modular State-Adaptive Landmark Tracking.  ... 
arXiv:2203.04275v3 fatcat:ba4wdwkxhjg7de7ic2h7wgd4qe

A Style and Semantic Memory Mechanism for Domain Generalization [article]

Yang Chen and Yu Wang and Yingwei Pan and Ting Yao and Xinmei Tian and Tao Mei
2021 arXiv   pre-print
We verify that it is critical for the network to be informative on what domain features are invariant and shared among instances, so that the network sharpens its understanding and improves its semantic  ...  Meanwhile, the inherent intra-domain style invariance is usually underappreciated and put on the shelf.  ...  , so that the network becomes Domain Adaptation (DA) algorithms aim to exploit agnostic to styles; or they rely on style decomposition ap- both annotated training data in the source domain  ... 
arXiv:2112.07517v1 fatcat:54ea7syojrcmhemdqcw6uhgs4a

Improving Transferability for Domain Adaptive Detection Transformers [article]

Kaixiong Gong, Shuang Li, Shugang Li, Rui Zhang, Chi Harold Liu, Qiang Chen
2022 arXiv   pre-print
DETR-style detectors stand out amongst in-domain scenarios, but their properties in domain shift settings are under-explored.  ...  The OTA module utilizes sliced Wasserstein distance to maximize the retention of location information while minimizing the domain gap in the decoder outputs.  ...  OAA provides domain-invariant features at pixel level while the OTA further reduces the domain gap at instance level.  ... 
arXiv:2204.14195v3 fatcat:ambs32a63renfjge5b6d46zb4q

A Domain Agnostic Normalization Layer for Unsupervised Adversarial Domain Adaptation [article]

Rob Romijnders, Panagiotis Meletis, Gijs Dubbelman
2018 arXiv   pre-print
As our normalization layer is domain agnostic at test time, we furthermore demonstrate that UADA using Domain Agnostic Normalization improves performance on unseen domains, specifically on Apolloscape  ...  Therefore, we propose a novel Domain Agnostic Normalization layer and thereby unlock the benefits of normalization layers for unsupervised adversarial domain adaptation.  ...  The domain classifier operates at the level of the representations, which might not suffice to reduce the domain adaptation gap.  ... 
arXiv:1809.05298v1 fatcat:3ymqfuvb6vaqtfpgjiuayycdzi

Generalizable Model-agnostic Semantic Segmentation via Target-specific Normalization [article]

Jian Zhang, Lei Qi, Yinghuan Shi, Yang Gao
2021 arXiv   pre-print
Concretely, we exploit the model-agnostic learning to simulate the domain shift problem, which deals with the domain generalization from the training scheme perspective.  ...  new coming) domains.  ...  Thus they aim to reduce the overfitting problem during training via the regularization mechanism. Carlucci et al .  ... 
arXiv:2003.12296v2 fatcat:t65rdkl6r5drhgiywrmer73qa4

Semi-supervised domain adaptation with CycleGAN guided by a downstream task loss [article]

Annika Mütze, Matthias Rottmann, Hanno Gottschalk
2022 arXiv   pre-print
semantic segmentation expert 2) the demonstration that the method is applicable to complex domain adaptation tasks and 3) a less biased domain gap analysis by using from scratch networks.  ...  Therefore, one would like to be able to train neural networks on synthetic domains, where data is abundant and labels are precise. However, these models often perform poorly on out-of-domain images.  ...  We then shift the out-of-domain input (real world) closer to the synthetic domain via a semisupervised image-to-image approach for mitigating the domain gap to the more abstract domain (e.g. simulation  ... 
arXiv:2208.08815v1 fatcat:bzcynov7dzacpblo5f6gcxzety

Bridge Segmentation Performance Gap Via Evolving Shape Prior

Chaoyu Chen, Xin Yang, Haoran Dou, Ruobing Huang, Xiaoqiong Huang, Xu Wang, Chong Duan, Shengli Li, Wufeng Xue, Pheng Ann Heng, Dong Ni
2020 IEEE Access  
In this paper, we propose a case adaptation strategy aiming to bridge the segmentation performance gap on domain-agnostic images. Our contribution is three-fold.  ...  However, it heavily depends on the expensive re-collection and re-training for domain-specific datasets and thus is not applicable to domain-agnostic images.  ...  ., imagespecific fine-tuning, is promising to bridge segmentation performance gaps on domain-agnostic images.  ... 
doi:10.1109/access.2020.3026073 fatcat:ennqhirmobah7ptdjvtec5d3qa

iFAN: Image-Instance Full Alignment Networks for Adaptive Object Detection

Chenfan Zhuang, Xintong Han, Weilin Huang, Matthew Scott
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
We propose Image-Instance Full Alignment Networks (iFAN) to tackle this problem by precisely aligning feature distributions on both image and instance levels: 1) Image-level alignment: multi-scale features  ...  fully exploited to establish a strong relationship among categories and domains.  ...  patch, and thus reduces the global image domain shift (e.g. image style, illumination, texture, etc.).  ... 
doi:10.1609/aaai.v34i07.7015 fatcat:j6iin27zu5gtrcx53gmsjlfcri

Adaptive Fine-Grained Sketch-Based Image Retrieval [article]

Ayan Kumar Bhunia, Aneeshan Sain, Parth Shah, Animesh Gupta, Pinaki Nath Chowdhury, Tao Xiang, Yi-Zhe Song
2022 arXiv   pre-print
., different drawing styles. Although this complicates the generalisation problem, fortunately, a handful of examples are typically available, enabling the model to adapt to the new category/style.  ...  To solve this new problem, we introduce a novel model-agnostic meta-learning (MAML) based framework with several key modifications: (1) As a retrieval task with a margin-based contrastive loss, we simplify  ...  [58] employed deep learning via deep triplet network to learn a common embedding space from heterogeneous domains.  ... 
arXiv:2207.01723v3 fatcat:b4dvtecwhvbvpjthpoo7jtw4jq

Semi-Supervised Domain Generalization in Real World: New Benchmark and Strong Baseline [article]

Luojun Lin, Han Xie, Zhifeng Yang, Zhishu Sun, Wenxi Liu, Yuanlong Yu, Weijie Chen, Shicai Yang, Di Xie
2022 arXiv   pre-print
To tackle this task, a straightforward solution is to propagate the class information from the labeled to the unlabeled domains via pseudo labeling in conjunction with domain confusion training.  ...  Yet, web data provides a free lunch to access a huge amount of unlabeled data with rich style information that can be harnessed to augment domain generalization ability.  ...  simultaneously via different DA methods, whilst testing the performance on the agnostic target domain.  ... 
arXiv:2111.10221v2 fatcat:ago4etkjn5bt3ap4bktnlvctzm
« Previous Showing results 1 — 15 out of 4,758 results