检索结果-内蒙古大学图书馆

Persistent Homology with Improved Locality Information for more Effective Delineation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Oner, Doruk Garin, Adélie Kozinski, Mateusz Hess, Kathryn Fua, Pascal The Computer Vision Laboratory Switzerland The Laboratory for Topology and Neuroscience École Polytechnique Fédérale de Lausanne Lausanne1015 Switzerland The Institute of Computer Graphics and Vision TU Graz Graz8010 Austria

Persistent Homology (PH) has been successfully used to train networks to detect curvilinear structures and to improve the topological quality of their results. However, existing methods are very global and ignore the location of topological features. In this paper, we remedy this by introducing a new filtration function that fuses two earlier approaches: thresholding-based filtration, previously used to train deep networks to segment medical images, and filtration with height functions, typically used to compare 2D and 3D shapes. We experimentally demonstrate that deep networks trained using our PH-based loss function yield reconstructions of road networks and neuronal processes that reflect ground-truth connectivity better than networks trained with existing loss functions based on PH. Code is available at https://***/doruk-oner/PH-TopoLoss. Copyright © 2021, The Authors. All rights reserved.

关键词： Topology

The Norm Must Go On: Dynamic Unsupervised Domain Adaptation by Normalization

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Mirza, Muhammad Jehanzeb Micorek, Jakub Possegger, Horst Bischof, Horst Institute for Computer Graphics and Vision Graz University of Technology Austria Christian Doppler Laboratory for Embedded Machine Learning Austria

Domain adaptation is crucial to adapt a learned model to new scenarios, such as domain shifts or changing data distributions. Current approaches usually require a large amount of labeled or unlabeled data from the shifted domain. This can be a hurdle in fields which require continuous dynamic adaptation or suffer from scarcity of data, e.g. autonomous driving in challenging weather conditions. To address this problem of continuous adaptation to distribution shifts, we propose Dynamic Unsupervised Adaptation (DUA). By continuously adapting the statistics of the batch normalization layers we modify the feature representations of the model. We show that by sequentially adapting a model with only a fraction of unlabeled data, a strong performance gain can be achieved. With even less than 1% of unlabeled data from the target domain, DUA already achieves competitive results to strong baselines. In addition, the computational overhead is minimal in contrast to previous approaches. Our approach is simple, yet effective and can be applied to any architecture which uses batch normalization. We show the utility of DUA by evaluating it on a variety of domain adaptation datasets and tasks including object recognition, digit recognition and object detection. © 2021, CC BY.

关键词： Object recognition

Optimization of ultrasonic imaging using persistence and filter technology

学校读者我要写书评

暂无评论

2013 2nd International Conference on Materials Science and Technology, ICMST 2013

作者： Guo, De Quan Yang, Hong Yu Zhang, Cong Yao Liu, Dong C. Key Laboratory of Fundamental Synthetic Vision Graphics and Image Science for National Defense Sichuan University Sichuan Province China Department of Machinery Engineering Neijiang Vocational and Technical College Neijiang China College of Computer Science Sichuan University Sichuan Province China

ISBN: (纸本)9783037857137

B-mode ultrasonic images are often pervaded by the electronics noise and speckle artifact, which may make the interpretation of medical images difficult. In this paper, a legible method for ultrasonic image is constructed and it can be effective to suppress the noise and improve the images quality. Dynamic persistence technology based on temporal multi-frames is utilized to smooth the electronics noise and even lower speckle artifact. To reduce computing complexity, a sum table method and conjugate direction search approaches are applied to speed up estimation of motion vector. The guided image filter strategy is used to enhance the contrast and detail resolution for the processed image by persistence. Experimental results show that the proposed method is effective to smooth noise and enhance contrast of ultrasound image in vivo. © (2013) Trans Tech Publications, Switzerland.

关键词： Ultrasonic imaging

SAILOR: Scaling Anchors via Insights into Latent Object Representation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Malić, Dušan Fruhwirth-Reisinger, Christian Possegger, Horst Bischof, Horst Institute of Computer Graphics and Vision Graz University of Technology Austria Christian Doppler Laboratory for Embedded Machine Learning Austria

LiDAR 3D object detection models are inevitably biased towards their training dataset. The detector clearly exhibits this bias when employed on a target dataset, particularly towards object sizes. However, object sizes vary heavily between domains due to, for instance, different labeling policies or geographical locations. State-of-the-art unsupervised domain adaptation approaches outsource methods to overcome the object size bias. Mainstream size adaptation approaches exploit target domain statistics, contradicting the original unsupervised assumption. Our novel unsupervised anchor calibration method addresses this limitation. Given a model trained on the source data, we estimate the optimal target anchors in a completely unsupervised manner. The main idea stems from an intuitive observation: by varying the anchor sizes for the target domain, we inevitably introduce noise or even remove valuable object cues. The latent object representation, perturbed by the anchor size, is closest to the learned source features only under the optimal target anchors. We leverage this observation for anchor size optimization. Our experimental results show that, without any retraining, we achieve competitive results even compared to state-of-the-art weakly-supervised size adaptation approaches. In addition, our anchor calibration can be combined with such existing methods, making them completely unsupervised. © 2022, CC BY.

关键词： Anchors

Training sequential on-line boosting classifier for visual tracking

学校读者我要写书评

暂无评论

Training sequential on-line boosting classifier for visual t...

International Conference on Pattern Recognition

作者： Helmut Grabner Jan Sochman Horst Bischof Jiri Matas Institute for Computer Graphics and Vision Graz University of Technology Austria Computer Vision Laboratory ETH Zurich Switzerland Center for Machine Perception Czech Technical University Prague Czech Republic

ISBN: (纸本)9781424421749

On-line boosting allows to adapt a trained classifier to changing environmental conditions or to use sequentially available training data. Yet, two important problems in the on-line boosting training remain unsolved: (i) classifier evaluation speed optimization and, (ii) automatic classifier complexity estimation. In this paper we show how the on-line boosting can be combined with Waldpsilas sequential decision theory to solve both of the problems. The properties of the proposed on-line WaldBoost algorithm are demonstrated on a visual tracking problem. The complexity of the classifier is changing dynamically depending on the difficulty of the problem. On average, a speedup of a factor of 5-10 is achieved compared to the non-sequential on-line boosting.

关键词： Boosting Decision theory computer graphics computer vision Training data Object detection Detectors Decision making Diversity reception Voting

Multicenter aortic vessel tree extraction using deep learning

学校读者我要写书评

暂无评论

Multicenter aortic vessel tree extraction using deep learnin...

Medical Imaging 2023: Biomedical Applications in Molecular, Structural, and Functional Imaging

作者： Scharinger, Bernhard Pepe, Antonio Jin, Yuan Gsaxner, Christina Li, Jianning Egger, Jan Graz University of Technology Institute for Computer Graphics and Vision Inffeldgasse 16c/II GrazA-8010 Austria Computer Algorithms for Medicine Laboratory Graz Austria Girardetstrasse 2 Essen45131 Germany Research Center for Connected Healthcare Big Data Zhejiang Lab Zhejiang Hangzhou311121 China

ISBN: (纸本)9781510660410

The aorta is the largest vessel of the human body and its pathological degenerations, such as dissections and aneurysms, can be life threatening. An automatic and fast segmentation of the aorta can therefore be a helpful tool to quickly identify an abnormal anatomy. The segmentation of the aortic vessel tree (AVT) typically requires extensive manual labor, but, in recent years, progress in deep learning techniques made the automation of this process viable. For this purpose, we tested different deep learning networks to segment the aortic vessel tree from computed tomography angiography (CTA) scans with a deep neural network consisting of an encoder-decoder architecture with skip connections and an optional self-attention block. The networks were trained on a dataset of 56 CTA scans from three different sources and resulted in Dice score similarities between 0.043-0.897. Generally, the classical U-Nets performed better than the ones containing a self-attention block, indicating that they might diminish performance for AVT segmentation. The quality of the resulting segmentations was highly dependent on the CTA image quality, especially on the contrast between the aorta and the surrounding tissues. However, the trained deep neural network can segment CTA scans well with limited computational resources and training data. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.

关键词： Deep neural networks

A two-stage Otsu'S thresholding based method on a 2D histogram

学校读者我要写书评

暂无评论

A two-stage Otsu'S thresholding based method on a 2D histogr...

IEEE International Conference on Intelligent computer Communication and Processing (ICCP)

作者： Puthipong Sthitpattanapongsa Thitiwan Srinark Graphics Innovation and Vision Engineering (GIVE) Laboratory Department of Computer Engineering Faculty of Engineering Kasetsart University Bangkok Thailand

We propose a fast and robust thresholding method that can overcome the shortcoming of the traditional and two-dimensional (2D) Otsu's methods. Our method takes advantage of the gradient gray level to divide the 2D histogram region, and applies the traditional Otsu's thresholding method twice on two projection histograms to separate the regions of interest from the background. The experimental results show that our method is robust against noise, and it requires less computational time than the 2D Otsu's and four other thresholding methods, which also give highly satisfactory and comparable results.

关键词： Histograms Magnetohydrodynamics Image segmentation Gaussian noise Robustness Shape

ActMAD: Activation Matching to Align Distributions for Test-Time-Training

学校读者我要写书评

暂无评论

ActMAD: Activation Matching to Align Distributions for Test-...

Conference on computer vision and Pattern Recognition (CVPR)

作者： M. Jehanzeb Mirza Pol Jané Soneira Wei Lin Mateusz Kozinski Horst Possegger Horst Bischof Institute for Computer Graphics and Vision TU Graz Austria Christian Doppler Laboratory for Embedded Machine Learning Institute of Control Systems KIT Germany Christian Doppler Laboratory for Semantic 3D Computer Vision

Test-Time-Training (TTT) is an approach to cope with out-of-distribution (OOD) data by adapting a trained model to distribution shifts occurring at test-time. We propose to perform this adaptation via Activation Matching (ActMAD): We analyze activations of the model and align activation statistics of the OOD test data to those of the training data. In contrast to existing methods, which model the distribution of entire channels in the ultimate layer of the feature extractor, we model the distribution of each feature in multiple layers across the network. This results in a more fine-grained supervision and makes ActMAD attain state of the art performance on CIFAR-100C and Imagenet-C. ActMAD is also architecture-and task-agnostic, which lets us go beyond image classification, and score 15.4% improvement over previous approaches when evaluating a KITTI-trained object detector on KITTI-Fog. Our experiments highlight that ActMAD can be applied to online adaptation in realistic scenarios, requiring little data to attain its full performance.

关键词：

ActMAD: Activation Matching to Align Distributions for Test-Time-Training

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Mirza, Muhammad Jehanzeb Soneira, Pol Jané Lin, Wei Kozinski, Mateusz Possegger, Horst Bischof, Horst Institute for Computer Graphics and Vision TU Graz Austria Christian Doppler Laboratory for Embedded Machine Learning Institute of Control Systems KIT Germany Christian Doppler Laboratory for Semantic 3D Computer Vision

Test-Time-Training (TTT) is an approach to cope with out-of-distribution (OOD) data by adapting a trained model to distribution shifts occurring at test-time. We propose to perform this adaptation via Activation Matching (ActMAD): We analyze activations of the model and align activation statistics of the OOD test data to those of the training data. In contrast to existing methods, which model the distribution of entire channels in the ultimate layer of the feature extractor, we model the distribution of each feature in multiple layers across the network. This results in a more fine-grained supervision and makes ActMAD attain state of the art performance on CIFAR-100C and Imagenet-C. ActMAD is also architecture- and task-agnostic, which lets us go beyond image classification, and score 15.4% improvement over previous approaches when evaluating a KITTI-trained object detector on KITTI-Fog. Our experiments highlight that ActMAD can be applied to online adaptation in realistic scenarios, requiring little data to attain its full performance. © 2022, CC BY.

关键词： Chemical activation