检索结果-内蒙古大学图书馆

Deep Learning vs. Traditional 3d Registration: A Featureless 3d Registration Baseline

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Bojanic, David Bartol, Kristijan Forest, Josep Petkovic, Tomislav Pribanic, Tomislav University of Zagreb Faculty of Electrical Engineering and Computing Unska 3 Zagreb10000 Croatia TU Dresden Dresden01069 Germany University of Girona Computer Vision and Robotics Research Institute Plaça de Sant Domènec 3 Girona17004 Spain

Recent 3D registration methods are mostly learning-based that either find correspondences in feature space and match them, or directly estimate the registration transformation from the given point cloud features. Therefore, these feature based methods have difficulties with generalizing onto point clouds that differ substantially from their training data. We propose a simple featureless traditional 3D registration baseline method based on the weighted cross-correlation between two given point clouds. Our method achieves strong results on current benchmarking datasets, outperforming most deep-learning methods. We further propose a methodology to create a 3D registration benchmark, given a point cloud dataset, that provides a more informative evaluation of a method w.r.t. other benchmarks. We also propose to use the 3D registration parameter range of a benchmark as an indicator of its difficulty. Using this methodology, we create a novel FAUST-partial (FP) benchmark, based on the FAUST dataset, with several difficulty levels. The FP benchmark addresses the limitations of the current benchmarks: lack of data and parameter range variability, and allows to evaluate the strengths and weaknesses of a 3D registration method w.r.t. a single registration parameter. Using the new FP benchmark, we provide a thorough analysis of the current state-of-the-art methods. Our source code is available on GitHub. © 2023, The Authors. All rights reserved.

关键词： Deep learning

Platform-portable reinforcement learning methods to localize underwater targets

学校读者我要写书评

暂无评论

Platform-portable reinforcement learning methods to localize...

OCEANS

作者： Ivan Masmitja Mario Martin Tom O’Reilly Narcis Palomeras Kakani Katija Joan Navarro Institut de Ciències del Mar (ICM) CSIC Barcelona Spain KEMLG Research Group Universitat Politècnica de Catalunya Barcelona Tech. Barcelona Spain Monterey Bay Aquarium Research Institute Moss Landing USA Computer Vision and Robotics Institute Universitat de Girona Girona Spain Bioinspiration Lab Monterey Bay Aquarium Research Institute Moss Landing USA

Subsea residential autonomous systems will reduce the cost and CO 2 footprint of future subsea operations. This concept includes subsea docking stations (SDS) and Autonomous Underwater Vehicles (AUVs). To ensure the AUVs are capable of interfacing with the SDS, a driftless and accurate close-range navigation schema is required. Thus, this paper proposes a visual method for navigation utilizing a defined set of ArUco markers printed on the SDS. To increase the accuracy of the system, the set of ArUco markers is defined to a common frame of reference, introducing more points in the camera frame for a more accurate solution to the Perspective-n-Point problem. This makes the system robust against partial occlusion. The paper has also implemented a modified Line of Sight (LOS) schema for the guidance of the vehicle. The results show stable pose estimates at up to 6 meters above the SDS, and successful autonomous docking of an underwater vehicle. The navigational estimates provided by the system get more stable when the vehicle is closer to the SDS.

关键词：

Exploiting Autoencoder's Weakness to Generate Pseudo Anomalies

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Astrid, Marcella Zaheer, Muhammad Zaigham Aouada, Djamila Lee, Seung-Ik Department of Artificial Intelligence University of Science and Technology Daejeon34113 Korea Republic of Field Robotics Research Section Electronics and Telecommunications Research Institute Daejeon34129 Korea Republic of Interdisciplinary Centre for Security Reliability and Trust University of Luxembourg Luxembourg1855 Luxembourg Department of Computer Vision Mohamed Bin Zayed University of Artificial Intelligence Abu Dhabi United Arab Emirates

Due to the rare occurrence of anomalous events, a typical approach to anomaly detection is to train an autoencoder (AE) with normal data only so that it learns the patterns or representations of the normal training data. At test time, the trained AE is expected to well reconstruct normal but to poorly reconstruct anomalous data. However, contrary to the expectation, anomalous data is often well reconstructed as well. In order to further separate the reconstruction quality between normal and anomalous data, we propose creating pseudo anomalies from learned adaptive noise by exploiting the aforementioned weakness of AE, i.e., reconstructing anomalies too well. The generated noise is added to the normal data to create pseudo anomalies. Extensive experiments on Ped2, Avenue, ShanghaiTech, CIFAR-10, and KDDCUP datasets demonstrate the effectiveness and generic applicability of our approach in improving the discriminative capability of AEs for anomaly detection. © 2024, CC BY-NC-SA.

关键词： Anomaly detection

Detection and depth estimation for domestic waste in outdoor environments by sensors fusion

学校读者我要写书评

暂无评论

arXiv 2022年

作者： de Páez-Ubieta, Ignacio L. Velasco-Sánchez, Edison Puente, Santiago T. Candelas, Francisco A. AUtomatics RObotics and Artificial Vision Lab Computer Science Research Institute University of Alicante San Vicente del Raspeig University Campus s/n San Vicente del Raspeig Alicante03690 Spain

In this work, we estimate the depth in which domestic waste are located in space from a mobile robot in outdoor scenarios. As we are doing this calculus on a broad range of space (0.3 - 6.0 m), we use RGB-D camera and LiDAR fusion. With this aim and range, we compare several methods such as average, nearest, median and center point, applied to those which are inside a reduced or non-reduced Bounding Box (BB). These BB are obtained from segmentation and detection methods which are representative of these techniques like Yolact, SOLO, You Only Look Once (YOLO)v5, YOLOv6 and YOLOv7. Results shown that, applying a detection method with the average technique and a reduction of BB of 40%, returns the same output as segmenting the object and applying the average method. Indeed, the detection method is faster and lighter in comparison with the segmentation one. The committed median error in the conducted experiments was 0.0298 ± 0.0544 m. © 2022, CC BY-NC-ND.

关键词： Mobile robots

Motion Planning and Control for Multi Vehicle Autonomous Racing at High Speeds

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Raji, Ayoub Liniger, Alexander Giove, Andrea Toschi, Alessandro Musiu, Nicola Morra, Daniele Verucchi, Micaela Caporale, Danilo Bertogna, Marko University of Modena and Reggio Emilia Italy University of Parma Italy Computer Vision Lab ETH Zurich Switzerland University of Pisa Italy Technology Innovation Institute Autonomous Robotics Research Center United Arab Emirates

This paper presents a multi-layer motion planning and control architecture for autonomous racing, capable of avoiding static obstacles, performing active overtakes, and reaching velocities above 75 m/s. The used offline global trajectory generation and the online model predictive controller are highly based on optimization and dynamic models of the vehicle, where the tires and camber effects are represented in an extended version of the basic Pacejka Magic Formula. The proposed single-track model is identified and validated using multi-body motorsport libraries which allow simulating the vehicle dynamics properly, especially useful when real experimental data are missing. The fundamental regularization terms and constraints of the controller are tuned to reduce the rate of change of the inputs while assuring an acceptable velocity and path tracking. The motion planning strategy consists of a Frenét-Frame-based planner which considers a forecast of the opponent produced by a Kalman filter. The planner chooses the collision-free path and velocity profile to be tracked on a 3 seconds horizon to realize different goals such as following and overtaking. The proposed solution has been applied on a Dallara AV-21 racecar and tested at oval race tracks achieving lateral accelerations up to 25 m/s2 Copyright © 2022, The Authors. All rights reserved.

关键词： Motion planning

Challenges of deep learning methods for COVID-19 detection using public datasets

学校读者我要写书评

暂无评论

Informatics in Medicine Unlocked 2022年 30卷 100945页

作者： Hasan, Md. Kamrul Alam, Md. Ashraful Dahal, Lavsen Roy, Shidhartho Wahid, Sifat Redwan Elahi, Md. Toufick E. Martí, Robert Khanal, Bishesh Department of Electrical and Electronic Engineering (EEE) Khulna University of Engineering & Technology (KUET) Khulna 9203 Bangladesh Nepal Applied Mathematics and Informatics Institute for Research (NAAMII) Nepal Computer Vision and Robotics Institute University of Girona Spain

Since the COVID-19 pandemic, several research studies have proposed Deep Learning (DL)-based automated COVID-19 detection, reporting high cross-validation accuracy when classifying COVID-19 patients from normal or other common Pneumonia. Although the reported outcomes are very high in most cases, these results were obtained without an independent test set from a separate data source(s). DL models are likely to overfit training data distribution when independent test sets are not utilized or are prone to learn dataset-specific artifacts rather than the actual disease characteristics and underlying pathology. This study aims to assess the promise of such DL methods and datasets by investigating the key challenges and issues by examining the compositions of the available public image datasets and designing different experimental setups. A convolutional neural network-based network, called CVR-Net (COVID-19 Recognition Network), has been proposed for conducting comprehensive experiments to validate our hypothesis. The presented end-to-end CVR-Net is a multi-scale-multi-encoder ensemble model that aggregates the outputs from two different encoders and their different scales to convey the final prediction probability. Three different classification tasks, such as 2-, 3-, 4-classes, are designed where the train–test datasets are from the single, multiple, and independent sources. The obtained binary classification accuracy is 99.8% for a single train–test data source, where the accuracies fall to 98.4% and 88.7% when multiple and independent train–test data sources are utilized. Similar outcomes are noticed in multi-class categorization tasks for single, multiple, and independent data sources, highlighting the challenges in developing DL models with the existing public datasets without an independent test set from a separate dataset. Such a result concludes a requirement for a better-designed dataset for developing DL tools applicable in actual clinical settings. The datas

关键词： Chest computed tomography and X-ray Convolutional neural networks COVID-19 disease Ensemble classifier

Multiparametric benthic landers for monitoring fishing-impacted deep-sea ecosystems

学校读者我要写书评

暂无评论

Multiparametric benthic landers for monitoring fishing-impac...

OCEANS

作者： Daniel Mihai Toma Jacopo Aguzzi Matias Carandell Marc Nogueras Enoc Martínez Marco Francescangeli Damianos Chatzievangelou Nixon Bahamon Joan Baptista Company Jordi Grinyo Marc Carreras Sascha Flögel Joaquín del Río SARTI Research Group Universitat Politècnica de Catalunya Vilanova i la Geltrú Spain Instituto de Ciencias del Mar Consejo Superior de Investigaciones Científicas Barcelona Spain Department of Ocean Systems NIOZ Royal Netherlands Institute for Sea Research Den Burg The Netherlands Computer Vision and Robotics Institute University of Girona Girona Spain GEOMAR Helmholtz Centre for Ocean Research Kiel Kiel Germany

To assess conservation efforts and follow the effects of anthropogenic pressures, it is crucial to monitor the ecological status of benthic habitats. Monitoring requires measurements that are made on-site, reproduced over a vast region, and capable of capturing their spatial evolution over time. This calls for the expensively efficient development of tools and protocols that deliver reliable, high-quality data. This article introduces the PLOME multiparametric benthic landers, a new tool and protocol for non-extractively assessing the status of fishes and benthic habitats. By combining independent and wirelessly connected benthic stations, AUVs, and USVs, PLOME creates a spatially adjustable, non-invasive, modular platform. Lander stations will be able to continuously deliver high-rate data while AUVs can provide comprehensive spatial measurements.

关键词：

The iToBoS dataset: skin region images extracted from 3D total body photographs for lesion detection

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Saha, Anup Adeola, Joseph Ferrera, Nuria Mothershaw, Adam Rezze, Gisele Gaborit, Séraphin D’Alessandro, Brian Hudson, James Szabó, Gyula Pataki, Balazs Rajani, Hayat Nazari, Sana Hayat, Hassan Primiero, Clare Soyer, H. Peter Malvehy, Josep Garcia, Rafael Computer Vision and Robotics Research Institute University of Girona Girona Spain Dermatology Department Hospital Clínic Barcelona Universitat de Barcelona Barcelona Spain Frazer Institute The University of Queensland Dermatology Research Center Brisbane Australia ISAHIT Paris France Canfield Scientific Inc. ParsippanyNJ United States V7 London United Kingdom HUN-REN Institute for Computer Science and Control Budapest Hungary CIBER de Enfermedades Raras Instituto de Salud Carlos III Barcelona Spain Dermatology Department Princess Alexandra Hospital BrisbaneQLD Australia Dermatology Department University of Trieste Trieste Italy

Artificial intelligence has significantly advanced skin cancer diagnosis by enabling rapid and accurate detection of malignant lesions. In this domain, most publicly available image datasets consist of single, isolated skin lesions positioned at the center of the image. While these lesion-centric datasets have been fundamental for developing diagnostic algorithms, they lack the context of the surrounding skin, which is critical for improving lesion detection. The iToBoS dataset was created to address this challenge. It includes 16,954 images of skin regions from 100 participants, captured using 3D total body photography. Each image roughly corresponds to a 7×9 cm section of skin with all suspicious lesions annotated using bounding boxes. Additionally, the dataset provides metadata such as anatomical location, age group, and sun damage score for each image. This dataset aims to facilitate training and benchmarking of algorithms, with the goal of enabling early detection of skin cancer and deployment of this technology in non-clinical environments. © 2025, CC BY-NC-SA.

关键词： Lung cancer

Cross Domain Object Detection by Target-Perceived Dual Branch Distillation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： He, Mengzhe Wang, Yali Wu, Jiaxi Wang, Yiru Li, Hanqing Li, Bo Gan, Weihao Wu, Wei Qiao, Yu ShenZhen Key Lab of Computer Vision and Pattern Recognition Shenzhen Institute of Advanced Technology Chinese Academy of Sciences China SenseTime Research University of Chinese Academy of Science China Shanghai AI Laboratory Shanghai China Beihang University China SIAT Branch Shenzhen Institute of Artificial Intelligence and Robotics for Society China

Cross domain object detection is a realistic and challenging task in the wild. It suffers from performance degradation due to large shift of data distributions and lack of instance-level annotations in the target domain. Existing approaches mainly focus on either of these two difficulties, even though they are closely coupled in cross domain object detection. To solve this problem, we propose a novel Target-perceived Dual-branch Distillation (TDD) framework. By integrating detection branches of both source and target domains in a unified teacher-student learning scheme, it can reduce domain shift and generate reliable supervision effectively. In particular, we first introduce a distinct Target Proposal Perceiver between two domains. It can adaptively enhance source detector to perceive objects in a target image, by leveraging target proposal contexts from iterative cross-attention. Afterwards, we design a concise Dual Branch Self Distillation strategy for model training, which can progressively integrate complementary object knowledge from different domains via self-distillation in two branches. Finally, we conduct extensive experiments on a number of widely-used scenarios in cross domain object detection. The results show that our TDD significantly outperforms the state-of-the-art methods on all the benchmarks. Our code and model will be available at here. Copyright © 2022, The Authors. All rights reserved.

关键词： Distillation