NewEvery arXiv paper, its researchers & institutions — mapped.
the archive

#adversarial training

13 results
cs.CV2019

Multi-level Domain Adaptive learning for Cross-Domain Detection

Rongchang Xie, Fei Yu, Jiachao Wang +2

The paper introduces a multi-level domain adaptive model that aligns both local and global feature distributions to improve object detection performance across different domains.

#domain adaptation#object detection#cross-domain detection#feature alignment
cs.CV2019

Towards Generating Stylized Image Captions via Adversarial Training

Omid Mohamad Nezami, Mark Dras, Stephen Wan +2

The paper presents ATTEND-GAN, an image captioning model that combines attention mechanisms with adversarial training to produce captions that are both content‑accurate and stylist…

#image captioning#stylized generation#adversarial training#attention mechanisms
cs.CV2019

Adversarial Seeded Sequence Growing for Weakly-Supervised Temporal Action Localization

Chengwei Zhang, Yunlu Xu, Zhanzhan Cheng +4

The paper proposes a weakly‑supervised framework for temporal action localization that alternates between a Seeded Sequence Growing network, which expands reliable seed regions, an…

#weakly-supervised learning#temporal action localization#adversarial training#seeded sequence growing
cs.CV2019

I Bet You Are Wrong: Gambling Adversarial Networks for Structured Semantic Segmentation

Laurens Samson, Nanne van Noord, Olaf Booij +3

The paper introduces a new adversarial training scheme for semantic segmentation where a "gambler" network bets on regions where the segmenter makes clear mistakes, encouraging the…

#semantic segmentation#adversarial training#structured prediction#uncertainty estimation
stat.ML2019

Robust Learning with Jacobian Regularization

Judy Hoffman, Daniel A. Roberts, Sho Yaida

The paper proposes an efficient Jacobian regularization method for neural networks that enlarges classification margins and improves robustness to both random and adversarial input…

#robustness#adversarial training#jacobian regularization#neural networks
eess.AS2019

An End-to-End Text-independent Speaker Verification Framework with a Keyword Adversarial Network

Sungrack Yun, Janghoon Cho, Jungyun Eum +2

The paper proposes an end-to-end speaker verification system that jointly trains a speaker embedding network and an automatic speech recognition network, using triplet loss and adv…

#speaker verification#text-independent#adversarial training#triplet loss
cs.SD2019

Adversarially Trained End-to-end Korean Singing Voice Synthesis System

Juheon Lee, Hyeong-Seok Choi, Chang-Bin Jeon +2

The paper introduces an end-to-end Korean singing voice synthesis system that converts lyrics and a symbolic melody into realistic singing audio using phonetic enhancement masking,…

#singing voice synthesis#end-to-end#adversarial training#phonetic masking
cs.CV2019

Semi-Supervised Adversarial Monocular Depth Estimation

Rongrong Ji, Ke Li, Yan Wang +6

The paper introduces a semi‑supervised adversarial framework that uses a small set of labeled image‑depth pairs together with many unlabeled monocular images to improve monocular d…

#monocular depth estimation#semi-supervised learning#adversarial training#depth prediction
cs.CV2019

Unsupervised Domain-Specific Deblurring via Disentangled Representations

Boyu Lu, Jun-Cheng Chen, Rama Chellappa

The paper proposes an unsupervised method for domain‑specific single‑image deblurring that separates content and blur features using disentangled encoders and leverages cycle‑consi…

#image deblurring#unsupervised learning#disentangled representations#domain‑specific restoration
eess.IV2019

Adversarially Trained Convolutional Neural Networks for Semantic Segmentation of Ischaemic Stroke Lesion using Multisequence Magnetic Resonance Imaging

Rachana Sathish, Ronnie Rajan, Anusha Vupputuri +2

The paper proposes an adversarially trained convolutional neural network to automatically segment core and penumbra regions of ischemic stroke lesions from multisequence MRI (DWI a…

#ischemic stroke segmentation#multisequence MRI#adversarial training#semantic segmentation
cs.CV2019

Moulding Humans: Non-parametric 3D Human Shape Estimation from Single Images

Valentin Gabeur, Jean-Sebastien Franco, Xavier Martin +2

The paper introduces a non‑parametric method that reconstructs full 3D human shape from a single RGB image by predicting a visible and a hidden depth map, which are combined like a…

#3d human shape estimation#single image reconstruction#depth map representation#adversarial training
cs.CV2019

Bilateral Adversarial Training: Towards Fast Training of More Robust Models Against Adversarial Attacks

Jianyu Wang, Haichao Zhang

The paper introduces Bilateral Adversarial Training, which jointly perturbs images and labels during training using a one-step targeted attack and a heuristic adversarial label, en…

#adversarial training#robustness#image classification#label perturbation
cs.CV2019

Joint Adversarial Training: Incorporating both Spatial and Pixel Attacks

Haichao Zhang, Jianyu Wang

The paper proposes a joint adversarial training approach that combines spatial transformation attacks with traditional pixel‑wise attacks to improve the robustness of deep vision m…

#adversarial training#spatial attacks#pixel attacks#model robustness