The YTU dataset and recurrent neural network based visual-inertial odometry

Gürtürk, Mert; Yusefi, Abdullah; Aslan, Muhammet Fatih; Soycan, Metin; Durdu, Akif; Masiero, Andrea

doi:10.1016/j.measurement.2021.109878

Cited by 18 publications

(17 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…VINet [175] 2017 Monocular + IMU CNN + LSTM Supervised VIOLearner [197] 2020 Monocular + IMU CNN Unsupervised VIO DeepVIO [187] 2019 Stereo + IMU CNN + LSTM Supervised Chen et al [188] 2019 Monocular + IMU FlowNet + LSTM Supervised Kim et al [198] 2021 Monocular + IMU CNN + LSTM Unsupervised Gurturk et al [199] 2021 Monocular + IMU CNN + LSTM Supervised…”

Section: Methods Year Sensor Neural Network Supervisionmentioning

confidence: 99%

An Overview on Visual SLAM: From Tradition to Semantic

et al. 2022

View full text Add to dashboard Cite

Visual SLAM (VSLAM) has been developing rapidly due to its advantages of low-cost sensors, the easy fusion of other sensors, and richer environmental information. Traditional visionbased SLAM research has made many achievements, but it may fail to achieve wished results in challenging environments. Deep learning has promoted the development of computer vision, and the combination of deep learning and SLAM has attracted more and more attention. Semantic information, as high-level environmental information, can enable robots to better understand the surrounding environment. This paper introduces the development of VSLAM technology from two aspects: traditional VSLAM and semantic VSLAM combined with deep learning. For traditional VSLAM, we summarize the advantages and disadvantages of indirect and direct methods in detail and give some classical VSLAM open-source algorithms. In addition, we focus on the development of semantic VSLAM based on deep learning. Starting with typical neural networks CNN and RNN, we summarize the improvement of neural networks for the VSLAM system in detail. Later, we focus on the help of target detection and semantic segmentation for VSLAM semantic information introduction. We believe that the development of the future intelligent era cannot be without the help of semantic technology. Introducing deep learning into the VSLAM system to provide semantic information can help robots better perceive the surrounding environment and provide people with higher-level help.

show abstract

Section: Methods Year Sensor Neural Network Supervisionmentioning

confidence: 99%

An Overview on Visual SLAM: From Tradition to Semantic

et al. 2022

View full text Add to dashboard Cite

show abstract

“…In these studies, methodological recommendations for SLAM were presented, and prediction errors in UAV poses were compared with previous studies. Indoor SLAM applications are often an important requirement in the robotics industry, such as search and rescue or defense in environments without GPS access, or in the entertainment industry, such as virtual reality [101,102]. However, as we know, there are no agricultural applications of such advanced SLAM methods in indoor environments such as greenhouses.…”

Section: Solution Proposal For Uav Applications In Greenhousesmentioning

confidence: 99%

A Comprehensive Survey of the Recent Studies with UAV for Precision Agriculture in Open Fields and Greenhouses

et al. 2022

Self Cite

View full text Add to dashboard Cite

The increasing world population makes it necessary to fight challenges such as climate change and to realize production efficiently and quickly. However, the minimum cost, maximum income, environmental pollution protection and the ability to save water and energy are all factors that should be taken into account in this process. The use of information and communication technologies (ICTs) in agriculture to meet all of these criteria serves the purpose of precision agriculture. As unmanned aerial vehicles (UAVs) can easily obtain real-time data, they have a great potential to address and optimize solutions to the problems faced by agriculture. Despite some limitations, such as the battery, load, weather conditions, etc., UAVs will be used frequently in agriculture in the future because of the valuable data that they obtain and their efficient applications. According to the known literature, UAVs have been carrying out tasks such as spraying, monitoring, yield estimation, weed detection, etc. In recent years, articles related to agricultural UAVs have been presented in journals with high impact factors. Most precision agriculture applications with UAVs occur in outdoor environments where GPS access is available, which provides more reliable control of the UAV in both manual and autonomous flights. On the other hand, there are almost no UAV-based applications in greenhouses where all-season crop production is available. This paper emphasizes this deficiency and provides a comprehensive review of the use of UAVs for agricultural tasks and highlights the importance of simultaneous localization and mapping (SLAM) for a UAV solution in the greenhouse.

show abstract

“…Recent neural network-based visual-inertial fusion schemes [2]- [6] all adopt a similar "feature fusion" scheme, where a neural network maps the raw (visual and inertial) measurements to a 6-DOF egomotion prediction in an end-to-end manner. Internally, the network extracts and combines (e.g., through concatenation of the two feature vectors) sensor-specific features.…”

Section: B Learning-based Approaches To Viomentioning

confidence: 99%

“…The resulting multimodal features pass into a final network component that predicts the egomotion. This scheme exists in both supervised [2]- [4] and self-supervised [5], [6] settings. In the self-supervised setting [16], a pixel-based reconstruction loss is minimized to jointly train a depth and egomotion network.…”

Section: B Learning-based Approaches To Viomentioning

confidence: 99%

See 1 more Smart Citation

A Self-Supervised, Differentiable Kalman Filter for Uncertainty-Aware Visual-Inertial Odometry

Wagstaff,

Wise,

Kelly

2022

Preprint

View full text Add to dashboard Cite

Traditionally, visual-inertial-odometry (VIO) systems rely on filtering or optimization-based frameworks for robot egomotion estimation. While these methods are accurate under nominal conditions, they are prone to failure in degraded environments, where illumination changes, fast camera motion, or textureless scenes are present. Learning-based systems have the potential to outperform classical implementations in degraded environments, but are, currently, less accurate than classical methods in nominal settings. A third class, of hybrid systems, attempts to leverage the advantages of both systems.Herein, we introduce a framework for training a hybrid VIO system. Our approach uses a differentiable Kalman filter with an IMU-based process model and a robust, neural networkbased relative pose measurement model. By utilizing the dataefficiency of self-supervised learning, we show that our system significantly outperforms a similar, supervised system, while enabling online retraining. To demonstrate the utility of our approach, we evaluate our system on a visually degraded version of the EuRoC dataset. Notably, we find that, in cases where classical estimators consistently diverge, our estimator does not diverge or suffer from a significant reduction in accuracy. Finally, our system, by properly utilizing the metric information contained in the IMU measurements, is able to recover metric scale, while other self-supervised monocular VIO approaches cannot.

show abstract

The YTU dataset and recurrent neural network based visual-inertial odometry

Cited by 18 publications

References 24 publications

An Overview on Visual SLAM: From Tradition to Semantic

An Overview on Visual SLAM: From Tradition to Semantic

A Comprehensive Survey of the Recent Studies with UAV for Precision Agriculture in Open Fields and Greenhouses

A Self-Supervised, Differentiable Kalman Filter for Uncertainty-Aware Visual-Inertial Odometry

Contact Info

Product

Resources

About