2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)最新文献

Robot Rotation Estimation Using Spherical Moments in Neural Networks 基于神经网络球矩的机器人旋转估计

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092205

Yao Du, C. Mateo, Omar Tahri

{"title":"Robot Rotation Estimation Using Spherical Moments in Neural Networks","authors":"Yao Du, C. Mateo, Omar Tahri","doi":"10.1109/AIPR57179.2022.10092205","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092205","url":null,"abstract":"This work proposes a method to estimate mobile robot rotation using spherical moments and machine learning. A polynomial fisheye camera model has been used to generate each image considering flying camera along a pre-planned trajectory. A dataset with 2400 images is created and labeled using Blender. The images are processed before it is fed into the neural network using 20 masks to reduces errors caused by non-common parts of the edge in image. afterwards, the spherical moments are obtained, as well as the triplets and the basis. Finally the basis is converted into a quaternion as input of network. Because for a neural network, a smaller number of inputs means faster training and higher accuracy. For the training phase, a strategy of data augmentation is applied. The data set is divided into many fragments (folders) according to a folder size that can be adjusted. Among them, in each training set, a sliding window is used to expend combinations within a certain range to calculate the rotation between two images. In tne end, the model is tested with synthesized fisheye camera data from Blender. The used simulated environment contains different light and weather conditions like fog or bad lighting situation. The proposed method in this paper is compared with different feature-based methods like FAST, ORB, SURF etc., which cannot successfully extract the feature points, to prove the performance of this novel approach. The results show that the method proposed in this paper can still accurately estimate","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"52 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122722238","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CNN-based network has Network Anisotropy -work harder to learn rotated feature than non-rotated feature 基于cnn的网络具有网络各向异性，学习旋转特征比学习非旋转特征更困难

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092224

Ashley Dale, Mei Qiu, Lauren A. Christopher, Wendy Krogg, Albert William

{"title":"CNN-based network has Network Anisotropy -work harder to learn rotated feature than non-rotated feature","authors":"Ashley Dale, Mei Qiu, Lauren A. Christopher, Wendy Krogg, Albert William","doi":"10.1109/AIPR57179.2022.10092224","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092224","url":null,"abstract":"Successful object identification and classification in a generic Convolutional Neural Network (CNN) depends on object orientation. We expect CNN-based architectures to work harder to learn a rotated version of a feature than when learning the same feature in its default orientation. We name this phenomenon “Network Anisotropy”. A data set of 6000 RGB and grayscale images was created with rotated orientations of a feature predetermined and evenly distributed across four classes: 0°, 30°, 60°, 90°. Four ResNet (18, 34, 50, 101) classifier architectures were trained and the confidence scores were used to represent prediction accuracy. The results show that in all networks, training performance lags several epochs for the 30° and 60° rotation predictions compared to the 0° and 90° rotations, indicating a quantifiable network anisotropy. Because 0° and 90° both lie along a single rectilinear axis that coincides with the convolutional kernel of the CNN, we expect the classifier to do better on these two classes than on 30° and 60° classes. This work confirms that CNN architectures may have weaker performance based on feature orientation alone, independent of the feature distribution within the data set or the correlation of features within an image.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127941012","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Latent-Space Analysis for Improved Overhead Imagery Explainability 改进架空图像可解释性的潜在空间分析

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092206

Joel Bornemann, Matthew D. Reisman, Steven Esposito, Timothy Berrill, Joshua Wall, Todd Conway, Jesse Boone, Ryan J. Soldin

{"title":"Latent-Space Analysis for Improved Overhead Imagery Explainability","authors":"Joel Bornemann, Matthew D. Reisman, Steven Esposito, Timothy Berrill, Joshua Wall, Todd Conway, Jesse Boone, Ryan J. Soldin","doi":"10.1109/AIPR57179.2022.10092206","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092206","url":null,"abstract":"With the ever-growing quantity of information generated by commercial and intelligence platforms, it is imperative that the community moves toward solutions that provide not just automated outputs, but a more accurate explanation into why automated outputs are predicted. Automated Target Recognition (ATR) solutions that rely on deep learning are being leveraged to derive intelligence automatically, but these networks can produce incorrect answers in some circumstances, reducing user trust. Most analysts cautiously trust these results, but they typically still require inefficient manual verification. To help address this challenge, we introduce an ATR anomaly metric created specifically for GEOINT systems that determines if a newly detected object fits into the distribution of the network’s original training dataset. This approach has been tested on different sets of targets, such as ships, airplanes, and vehicles, spanning multiple publicly available EO and SAR commercial GEOINT datasets. Low anomaly scores generated by this approach help to validate model performance and increase trust in a model’s outputs, while high anomaly scores help identify unique images that were not captured within a model’s training distribution. For instance, some examples of high anomaly scores were found to correspond to mislabeled data used in training, identifying aircraft in pre/post flight configurations and in various stages of disassembly, detecting boats out of water, and identifying aircraft that were viewed through clouds. More importantly, for detections with a low anomaly score, our analysis shows that the results are highly accurate, allowing the analyst to trust the results and focus on harder problems. Additionally, we show a preliminary proof of concept analysis in which we invert this anomaly metric to perform semi-supervised active learning on brand new datasets.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124373417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 1

SMU-DDI Cyber Autonomy Range SMU-DDI网络自治范围

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092228

Darrell L. Young, Mark Bigham, M. Bradbury, Eric D. Larson, Mitchell A. Thornton

{"title":"SMU-DDI Cyber Autonomy Range","authors":"Darrell L. Young, Mark Bigham, M. Bradbury, Eric D. Larson, Mitchell A. Thornton","doi":"10.1109/AIPR57179.2022.10092228","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092228","url":null,"abstract":"The Southern Methodist University-Darwin Deason Institute for Cybersecurity (SMU-DDI) Cyber Autonomy Range (CAR) addresses the incorporation of increased resiliency, reliability, and cyber security of the autonomous systems (AS) cyberinfrastructure; an issue with widespread concern and broad impact on society. The advances of data science and Machine Learning/Artificial Intelligence (ML/AI) methods coupled with their integration into autonomous subsystems is an enabling trend that supports AS maturity. Likewise, these same aspects of ML/AI present entirely new aspects of cyber security, many of which have only been analyzed in a preliminary sense or for special cases. The ML/AI aspects of cyber security are critically important, with significant ramifications in human safety and well-being. The CAR is a collaborative facility that supports the assessment of AS when faced with cyber threats by assessing their attack surface, vulnerability, and their degree of resistance to such threats. It is instrumented to simulate and/or emulate the external environment of an AS and can subject the AS to a variety of controlled cyber-attacks. Because the decision-making capabilities of many AS are based upon data-driven ML/AI-enabled technologies, the threat surface surrounding ML/AI subsystems is of particular concern. The CAR is especially configured to investigate and simulate (or emulate) cyber-attacks on ML/AI-equipped subsystems; particularly ML/AI subsystems that depend upon data sources derived from sensor suites or other data sources.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115990166","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 2

Peanut Seed Germination Detection from Aerial Images 基于航空图像的花生种子萌发检测

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092219

Shengfang Ma, Yuting Zhou, K. C. Flynn, Sathyanarayanan N. Aakur

{"title":"Peanut Seed Germination Detection from Aerial Images","authors":"Shengfang Ma, Yuting Zhou, K. C. Flynn, Sathyanarayanan N. Aakur","doi":"10.1109/AIPR57179.2022.10092219","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092219","url":null,"abstract":"Peanut is an essential economic oil crop around the world. Therefore, accurate and real-time detection of peanut seed germination is necessary for peanut field management. However, traditional peanut seedlings’ germination monitoring is time-consuming and labor intense, especially for large fields. In this work, we propose to reduce the time lag in detecting peanut germination failures by combining the power of deep learning-based object detection (OD) and unmanned aerial systems (UAS) to identify early in-field peanut germination. To find the most suitable object detection model, we first compared the performance of two representative OD models, Faster RCNN and SSD, to identify peanut seedlings from UAS imagery obtained through a multispectral camera setup (MicaSense Rededge). The results showed that the F1 score of the SSD model is 0.82, while it is 0.85 for Faster RCNN at an Intersection over Union (IoU) of 0.5. Through extensive ablations, we find that deeper models only marginally improved the performance but were more expensive in computation and inference times. Interestingly we find that the performance of RGB-based seedling detection (0.917 mAP) is comparable to that of R-RedEdge-NIR (0.919 mAP), indicating that a remote sensing setup with a regular RGB camera can perform as well as a more expensive, multispectral camera system to detect peanut seedlings. With extensive experimentation, we infer that cheaper remote sensing mechanisms with the rapid acquisition of UAS-based imagery and the efficiency of OD methods are practical for early peanut germination detection.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127214100","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploiting Simplified Depth Estimation for Stereo-based 2D Object Detection 利用简化的深度估计进行立体二维目标检测

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092234

Yaesop Lee, Hyungtae Lee, E. Lee, H. Kwon, S. Bhattacharyya

{"title":"Exploiting Simplified Depth Estimation for Stereo-based 2D Object Detection","authors":"Yaesop Lee, Hyungtae Lee, E. Lee, H. Kwon, S. Bhattacharyya","doi":"10.1109/AIPR57179.2022.10092234","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092234","url":null,"abstract":"Stereo image inputs provide higher object detection accuracy than monocular images by enabling the detection of objects that are missed from one view while being detectable from another view. To take advantage of additional information from the secondary image, it is necessary to search for the corresponding region in the images of different views by projecting with depth information of the target object. However, most existing studies utilize highly complex computations to estimate the depth for simple 2D object detection. This complexity limits the potential for deploying the methods on platforms, such as unmanned aerial vehicles, that involve significant resource constraints. In this paper, we introduce a simplified depth approximation to obtain depth information by quantizing the depth values into a small number of representative values. With these values, the regions of interest are projected to the secondary image to concatenate the information from the additional image. We validate our method with the KITTI dataset. Our results show that while having very low complexity, our approximation method leads to greatly improved object detection performance in two out of three difficulty groups of the dataset, and comparable performance in the other difficulty group compared to use of monocular image input.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"35 5","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113935953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Utilization of Artificial Intelligence in Automated Image Analysis 人工智能在自动图像分析中的应用

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092210

Nicholas S. Noboa, John W. Von Holle, Paul A. Brown, J. Irvine

引用次数: 0

What Does Learning About Time Tell About Outdoor Scenes? 了解时间告诉我们户外场景的什么?

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092235

Zeyu Zhang, Callista Baker, Noor Azam-Naseeruddin, Jingzhou Shen, Robert Pless

引用次数: 0

Informing selection of performance metrics for medical image segmentation evaluation using configurable synthetic errors 使用可配置的合成误差为医学图像分割评估提供性能指标的选择

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092203

Shuyue Guan, Ravi K. Samala, Weijie Chen

{"title":"Informing selection of performance metrics for medical image segmentation evaluation using configurable synthetic errors","authors":"Shuyue Guan, Ravi K. Samala, Weijie Chen","doi":"10.1109/AIPR57179.2022.10092203","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092203","url":null,"abstract":"Machine learning-based segmentation in medical imaging is widely used in clinical applications from diagnostics to radiotherapy treatment planning. Segmented medical images with ground truth is useful to investigate the properties of different segmentation performance metrics to inform metric selection. Regular geometrical shapes are often used to synthesize segmentation errors and illustrate properties of performance metrics, but they lack the complexity of anatomical variations in real images. In this study, we present a tool to emulate segmentations by adjusting the reference (truth) masks of anatomical objects extracted from real medical images. Our tool is designed to modify the defined truth contours and emulate different types of segmentation errors with a set of user-configurable parameters. We defined the ground truth objects from 230 patient images in the Glioma Image Segmentation for Radiotherapy (GLIS-RT) database. For each object, we used our segmentation synthesis tool to synthesize 10 versions of segmentation (i.e., 10 simulated segmentors or algorithms) where each version has a pre-defined combination of segmentation errors. We then applied 20 performance metrics to evaluate all synthetic segmentations. We demonstrated the properties of these metrics including their ability to capture specific types of segmentation errors. By analyzing the intrinsic properties of these metrics and categorizing the segmentation errors, we are working toward the goal of developing a decision-tree tool for assisting selection of segmentation performance metrics.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130111459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Traversability Assessment and Reachability Analysis for Unmanned Ground Vehicles through Image Segmentation 基于图像分割的无人地面车辆可穿越性评估与可达性分析

2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) Pub Date : 2022-10-11 DOI: 10.1109/AIPR57179.2022.10092223

Philip Bailey, Andrew Clevenger, H. Sevil

引用次数: 0