Filters








8,004 Hits in 4.7 sec

Object Segmentation using Pixel-wise Adversarial Loss [article]

Ricard Durall, Franz-Josef Pfreundt, Ullrich Köthe, Janis Keuper
2019 arXiv   pre-print
The key element of our method is to replace the commonly used binary adversarial loss with a high resolution pixel-wise loss.  ...  We show, that this combination of pixel-wise adversarial training and weight averaging leads to significant and consistent gains in segmentation performance, compared to the baseline models.  ...  In this paper, we address the task of object segmentation by proposing an adversarial learning scheme which adds a pixel-wise adversarial loss to the classical topology.  ... 
arXiv:1909.10341v1 fatcat:yp7sotlz5zgqjelotztutkd6r4

Adaptive Affinity Fields for Semantic Segmentation [article]

Tsung-Wei Ke, Jyh-Jing Hwang, Ziwei Liu, Stella X. Yu
2018 arXiv   pre-print
Semantic segmentation has made much progress with increasingly powerful pixel-wise classifiers and incorporating structural priors via Conditional Random Fields (CRF) or Generative Adversarial Networks  ...  We use adversarial learning to select the optimal affinity field size for each semantic category.  ...  Summary We propose adaptive affinity fields (AAF) for semantic segmentation, which incorporate geometric regularities into segmentation models, and learn local relations with adaptive ranges through adversarial  ... 
arXiv:1803.10335v3 fatcat:2o74jzivonbarfafrvhfyeqopu

Adaptive Affinity Fields for Semantic Segmentation [chapter]

Tsung-Wei Ke, Jyh-Jing Hwang, Ziwei Liu, Stella X. Yu
2018 Lecture Notes in Computer Science  
Semantic segmentation has made much progress with increasingly powerful pixel-wise classifiers and incorporating structural priors via Conditional Random Fields (CRF) or Generative Adversarial Networks  ...  We use adversarial learning to select the optimal affinity field size for each semantic category.  ...  Summary We propose adaptive affinity fields (AAF) for semantic segmentation, which incorporate geometric regularities into segmentation models, and learn local relations with adaptive ranges through adversarial  ... 
doi:10.1007/978-3-030-01246-5_36 fatcat:icz2qozfrng57ocxi6as5vydlu

No More Discrimination: Cross City Adaptation of Road Scene Segmenters [article]

Yi-Hsin Chen, Wei-Yu Chen, Yu-Ting Chen, Bo-Cheng Tsai, Yu-Chiang Frank Wang, Min Sun
2017 arXiv   pre-print
By advancing a joint global and class-specific domain adversarial learning framework, adaptation of pre-trained segmenters to that city can be achieved without the need of any user annotation or interaction  ...  By utilizing Google Street View and its time-machine feature, we can collect unannotated images for each road scene at different times, so that the associated static-object priors can be extracted accordingly  ...  domain adversarial loss, respectively.  ... 
arXiv:1704.08509v1 fatcat:7ldqict6yzba5cm6j2aqsdqcu4

SegAN: Adversarial Network with Multi-scale L1 Loss for Medical Image Segmentation

Yuan Xue, Tao Xu, Han Zhang, L. Rodney Long, Xiaolei Huang
2018 Neuroinformatics  
Instead, we use a fully convolutional neural network as the segmentor to generate segmentation label maps, and propose a novel adversarial critic network with a multi-scale L_1 loss function to force the  ...  We tested our SegAN method using datasets from the MICCAI BRATS brain tumor segmentation challenge.  ...  [9] used both the conventional adversarial loss of GAN and pixel-wise softmax loss against ground truth.  ... 
doi:10.1007/s12021-018-9377-x pmid:29725916 fatcat:njcwonnnhzgmnhjydsoss7hpee

SROBB: Targeted Perceptual Loss for Single Image Super-Resolution [article]

Mohammad Saeed Rad, Behzad Bozorgtabar, Urs-Viktor Marti, Max Basler, Hazim Kemal Ekenel, Jean-Philippe Thiran
2019 arXiv   pre-print
In particular, the proposed method leverages our proposed OBB (Object, Background and Boundary) labels, generated from segmentation labels, to estimate a suitable perceptual loss for boundaries, while  ...  In this paper, we propose a novel method to benefit from perceptual loss in a more objective way.  ...  The MSE and adversarial loss terms are defined as follows: • Pixel-wise loss It is by far the most commonly used loss function in SR.  ... 
arXiv:1908.07222v1 fatcat:ygnwxgjpwnctnb4omufprrdpt4

Defending From Physically-Realizable Adversarial Attacks Through Internal Over-Activation Analysis [article]

Giulio Rossolini, Federico Nesti, Fabio Brau, Alessandro Biondi, Giorgio Buttazzo
2022 arXiv   pre-print
The presented defense relies on specific Z-score analysis performed on the internal network features to detect and mask the pixels corresponding to adversarial objects in the input image.  ...  The effectiveness of Z-Mask is evaluated with an extensive set of experiments carried out on models for both semantic segmentation and object detection.  ...  In practice, a pixel-wise max function is used for F(•).  ... 
arXiv:2203.07341v1 fatcat:ti6xsuxybbabba2pldcngfjflq

Unsupervised Image Segmentation by Mutual Information Maximization and Adversarial Regularization [article]

S. Ehsan Mirsadeghi, Ali Royat, Hamid Rezatofighi
2021 arXiv   pre-print
To customize an adversarial training scheme for the problem, we incorporate adversarial pixel noise along with spatial perturbations to impose photometrical and geometrical invariance on the deep neural  ...  Our experiments demonstrate that our method achieves the state-of-the-art performance on two commonly used unsupervised semantic segmentation datasets, COCO-Stuff, and Potsdam.  ...  pixel-wise label predictions.  ... 
arXiv:2107.00691v1 fatcat:y3wgpxrpqfapnk4otstxebc6pa

ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation

Tuan-Hung Vu, Himalaya Jain, Maxime Bucher, Matthieu Cord, Patrick Perez
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
In this work, we address the task of unsupervised domain adaptation in semantic segmentation with losses based on the entropy of the pixel-wise predictions.  ...  To this end, we propose two novel, complementary methods using (i) an entropy loss and (ii) an adversarial loss respectively.  ...  A similar approach of global and class-wise alignment is used in [5] with the class-wise alignment being done using adversarial training on grid-wise soft pseudolabels.  ... 
doi:10.1109/cvpr.2019.00262 dblp:conf/cvpr/VuJBCP19 fatcat:ywxufgupfvfejjfuzngweqcbeq

Unsupervised Domain Adaption for High-Resolution Coastal Land Cover Mapping with Category-Space Constrained Adversarial Network

Jifa Chen, Guojun Zhai, Gang Chen, Bo Fang, Ping Zhou, Nan Yu
2021 Remote Sensing  
Additionally, the low inter-class variances and intricate spatial details of coastal objects may entail poor presentation.  ...  Coastal land cover mapping (CLCM) across image domains presents a fundamental and challenging segmentation task.  ...  Alternate adversarial training is driven by the objective function L, in which the segmentation loss L S seg and adversarial loss L i adv , respectively, serve for the dense prediction and multi-level  ... 
doi:10.3390/rs13081493 fatcat:j2vpmd4smrcvvmpszk43bvt52a

ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation [article]

Tuan-Hung Vu, Himalaya Jain, Maxime Bucher, Matthieu Cord, Patrick Pérez
2019 arXiv   pre-print
In this work, we address the task of unsupervised domain adaptation in semantic segmentation with losses based on the entropy of the pixel-wise predictions.  ...  To this end, we propose two novel, complementary methods using (i) entropy loss and (ii) adversarial loss respectively.  ...  A similar approach of global and class-wise alignment is used in [5] with the class-wise alignment being done using adversarial training on grid-wise soft pseudolabels.  ... 
arXiv:1811.12833v2 fatcat:e7ox63x7svbzvlvnalfbleotp4

Adversarial Policy Gradient for Deep Learning Image Augmentation [article]

Kaiyang Cheng, Claudia Iriondo, Francesco Calivá, Justin Krogue, Sharmila Majumdar, Valentina Pedoia
2019 arXiv   pre-print
The use of semantic segmentation for masking and cropping input images has proven to be a significant aid in medical imaging classification tasks by decreasing the noise and variance of the training dataset  ...  Our method, Adversarial Policy Gradient Augmentation (APGA), shows promising results on Stanford's MURA dataset and on a hip fracture classification task with an increase in global accuracy of up to 7.33%  ...  To mask out the image-level features that are less useful for the classification task, we use the segmentation model to produce the pixel-wise probability p k of the pixel being useful.  ... 
arXiv:1909.04108v1 fatcat:g2kxhgz2p5fk7bru5rvaohwfde

Structured Knowledge Distillation for Dense Prediction [article]

Yifan Liu, Changyong Shun, Jingdong Wang, Chunhua Shen
2020 arXiv   pre-print
Specifically, we study two structured distillation schemes: i) pair-wise distillation that distills the pair-wise similarities by building a static graph; and ii) holistic distillation that uses adversarial  ...  The effectiveness of our knowledge distillation approaches is demonstrated by experiments on three dense prediction tasks: semantic segmentation, depth estimation and object detection.  ...  Optimization The overall objective function consists of a standard multiclass cross-entropy loss mc (S) with pixel-wise and structured distillation terms 1 (S, D) = mc (S) + λ 1 ( pi (S) + pa (S)) − λ  ... 
arXiv:1903.04197v7 fatcat:vcwpcgffgndwfo3xye3kdjtm24

3D Scene Parsing via Class-Wise Adaptation [article]

Daichi Ono, Hiroyuki Yabe, Tsutomu Horikawa
2019 arXiv   pre-print
Especially in the case of semantic segmentation, annotating pixel by pixel takes a significant amount of time and often makes mistakes.  ...  We propose the method that uses only computer graphics datasets to parse the real world 3D scenes. 3D scene parsing based on semantic segmentation is required to implement the categorical interaction in  ...  Pixel-wise loss which uses local information for domain adaptation helps for local domain shift. Figure 2 shows the training procedure.  ... 
arXiv:1812.03622v2 fatcat:f3p3x42u5vcfre4d7tooh2hkoq

Pixel-by-Pixel Cross-Domain Alignment for Few-Shot Semantic Segmentation [article]

Antonio Tavera, Fabio Cermelli, Carlo Masone, Barbara Caputo
2021 arXiv   pre-print
We propose a novel pixel-by-pixel domain adversarial loss following three criteria: (i) align the source and the target domain for each pixel, (ii) avoid negative transfer on the correctly represented  ...  In this context, aligning the domains is made more challenging by the pixel-wise class imbalance that is intrinsic in the segmentation and that leads to ignoring the underrepresented classes and overfitting  ...  To do this, we introduce the Pixel-By-Pixel Cross-Domain Alignment framework (PixDA), that uses a novel pixel-wise discriminator and modulates the adversarial loss for each pixel to: (i) align pixel-wise  ... 
arXiv:2110.11650v1 fatcat:b2gusgf7pfgktmwgldvkf7btti
« Previous Showing results 1 — 15 out of 8,004 results