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

{"title":"Increasing Trust in Artificial Intelligence with a Defensible AI Technique","authors":"J. Straub","doi":"10.1109/AIPR57179.2022.10092226","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092226","url":null,"abstract":"Artificial intelligence (AI) systems are prevalent throughout society and have significant impact on peoples’ everyday lives. Explainable artificial intelligence techniques have been developed in response to concerns about the opaqueness of many AI techniques, given their potential impact. Explainable techniques, though, do not guarantee performance – instead, they simply explain what was recommended or decided. This paper discusses how a new class of AI techniques, defensible AI (DAI) can increase public trust in AI. Issues with existing techniques are reviewed and an example DAI technique is presented and its benefits – relative to reliability, trust and security – are assessed.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"13 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":"126684692","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}

{"title":"Codomain Transformation to Represent Abstract Information for Machine Learning in Visual Scenes","authors":"A. Kalukin","doi":"10.1109/AIPR57179.2022.10092232","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092232","url":null,"abstract":"Convolutional Neural Networks (CNNs) succeed well in learning to recognize spatial and spectral image features which are easily distinguished by convolutional operations; for example, visual cues based on shape and color of isolated objects. However, they are challenged in learning features which are rule-based and abstract; for example, the number of elements in a scene (cardinality), and other relationships within a scene that have to do with subtle criteria of order and relative position, especially when essential cues occupy a small fraction of the spatial and spectral search space and can vary significantly in appearance. The computational experiment reported in this paper demonstrates the efficacy of codomain transformation to help CNNs recognize the abstract concept of cardinality for artificially generated geometric objects of various shapes, sizes, and colors.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"05 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":"130803655","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}

{"title":"Trusted Data Anomaly Detection (TaDA) in Ground Truth Image Data","authors":"William Boler, Ashley Dale, Lauren A. Christopher","doi":"10.1109/AIPR57179.2022.10092217","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092217","url":null,"abstract":"Current state-of-the-art Artificial Intelligence (AI) anomaly detection from images is primarily used for defect detection and relies on relatively homogeneous datasets of images with similar foregrounds and backgrounds. This type of anomaly detection uses human labelled ground-truth data. In our research, we have extremely heterogeneous datasets and want to identify outliers. We use self-supervised Variational Autoencoders (VAEs) to identify anomalies in the latent vector feature space. Understanding the outliers in a large training data set is important for establishing trustworthiness of the AI models learned from these data, a strong requirement for military AI applications. Our study uses 8984 examples from Kaggle military planes and 4300 examples from Kaggle landscape data. We present the results of the combined heterogeneous dataset on the localized methods, with one such result exhibiting inliers as landscapes/backgrounds and outliers as all aircraft, detecting aircraft as anomalies with a 0.87 AUC. Results also include the inter-class AUC across the different aircraft classes. Our contribution to the state-of-the-art is to apply isolation forests to the latent space data after UMAP embeddings in a strongly heterogeneous image dataset for military applications to identify anomalies.","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":"127108328","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}

{"title":"An Intelligent Traffic Monitoring Embedded System using Video Data Mining","authors":"Maria J. P. Peixoto, Akramul Azim, Jim Sheehan, D. Timothy","doi":"10.1109/AIPR57179.2022.10092207","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092207","url":null,"abstract":"Traffic monitoring through video analysis facilitates understanding the daily driving profile, allowing action to be taken in operational deviations and mitigating the accidents risks or incidents resulting from human error, offering a safer operation for the population. This paper discusses the workflow to create a predictive traffic monitoring system using video data mining on embedded systems. Our proposal uses machine learning, classification models, and tracking techniques to efficiently and accurately extract information about vehicle type, direction, dimensions, and speed through image capture and analysis. The analysis will provide a means to forecast trends using current and historical data and is also intended to be used in conjunction with the NVIDIA JetPack SDK to lower the costs of installing and maintaining sensors. We show that it is possible to classify and track vehicles using artifacts at a lower cost than those already used and without having to interdict the road for sensor structuring.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"9 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":"115321670","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}

{"title":"An Evaluation of Federated Learning Techniques for Secure and Privacy-Preserving Machine Learning on Medical Datasets","authors":"Abdulkadir Korkmaz, Ahmad Alhonainy, P. R. Rao","doi":"10.1109/AIPR57179.2022.10092212","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092212","url":null,"abstract":"Federated learning is a machine learning technique that trains a machine learning model across multiple decentralized edge devices or servers holding local data samples without revealing them. It is a technique that ensures the privacy of data stored (locally on a server/edge device) while training a machine learning model. This technique is particularly valuable for hospitals to collaboratively learn machine learning models without moving all the training data (e.g., patient records) to a centralized location due to federal laws. In this experimental study, we investigate how federated learning performs on different real-world medical datasets using a cluster of machines. We tested different deep learning models and federated optimization strategies on CloudLab, which is an experimental testbed for networking and distributed computing research. Our experiments showed that Inception-v3 and EfficientNetB0 were usually the top two performing models based on accuracy computed over the test set. For federated optimizations, FedAvg was the best strategy followed by FedAvgM.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"19 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":"116471417","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}

{"title":"A comparison of features Synthetic WAMI and GES of the same location","authors":"Koundinya Nouduri, Jason James, Joshua Fraser, H. Akbarpour, K. Palaniappan","doi":"10.1109/AIPR57179.2022.10092236","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092236","url":null,"abstract":"The rapid growth in deep learning has accelerated advances in many areas of computer vision. However, deep learning-based approaches require a large amount of data to train models. Subsequently, synthetic data is increasingly being looked to as a source for labeled training datasets to be used with supervised deep learning algorithms. WAMI (wide-area motion imagery) is sequential, oblique imagery, typically taken from an aircraft or drone, at city scale. Collecting a WAMI dataset can represent a significant investment of resources and logistics. However, the availability of synthetic WAMI datasets could overcome these concerns as well as potentially add benefits such as having associated ground truth. Recently, Google released Earth Studio, a browser-based animation tool that uses a 3D rendering engine to generate WAMI-like datasets across the globe. When working with synthetic data, a key point of concern is whether the synthetic data is sufficiently realistic for the purpose at hand. In this paper, we generate rendered WAMI datasets using Google Earth Studio. The rendered datasets are of the same locations for which we also have real WAMI datasets, and we then analyze the WAMI dataset and the images rendered by Google Earth Studio based on 3D reconstruction and feature evaluation of the dataset to determine how feasible synthetic datasets are in comparison to non-synthetic ones.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"24 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":"127769730","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}

{"title":"Is the Next Winter Coming for AI? Elements of Making Secure and Robust AI","authors":"Joshua D. Harguess, C. Ward","doi":"10.1109/AIPR57179.2022.10092230","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092230","url":null,"abstract":"While the recent boom in Artificial Intelligence (AI) has given rise to the technology's use and popularity across many domains, the same boom has exposed vulnerabilities of the technology to many threats that could cause the next \"AI winter\". AI is no stranger to \"winters\", or drops in funding and interest in the technology and its applications. Many in the field consider the early 1970's as the first AI winter with another proceeding in the late 1990's and early 2000's. There is some consensus that another AI winter is all but inevitable in some shape or form, however, current thoughts on the next winter do not consider secure and robust AI and the implications of the success or failure of these areas. The emergence of AI as an operational technology introduces potential vulnerabilities to AI's longevity. The National Security Commission on AI (NSCAI) report outlines recommendations for building secure and robust AI, particularly in government and Department of Defense (DoD) applications. However, are they enough to help us fully secure AI systems and prevent the next \"AI winter\"? An approaching \"AI Winter\" would have a tremendous impact in DoD systems as well as those of our adversaries. Understanding and analyzing the potential of this event would better prepare us for such an outcome as well as help us understand the tools needed to counter and prevent this \"winter\" by securing and robustifying our AI systems. In this paper, we introduce the following four pillars of AI assurance, that if implemented, will help us to avoid the next AI winter: security, fairness, trust, and resilience.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"33 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":"123932905","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}

{"title":"Attention Mechanism Improves YOLOv5x for Detecting Vehicles on Surveillance Videos","authors":"Mei Qiu, Lauren A. Christopher, Stanley Y. P. Chien, Yaobin Chen","doi":"10.1109/AIPR57179.2022.10092237","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092237","url":null,"abstract":"Vehicle detection accuracy on surveillance videos is heavily restricted by camera angles, low lighting conditions, low visibility caused by harsh weather, and serious occlusions. For the full 24/7 operation, the Intelligent Transportation Services (ITS) are expected to perform well on all the categories of the target detections in the environment. Unfortunately, most existing datasets do not cover all these difficult conditions. Moreover, the state-of-the-art Deep Learning detector performance decreases for these difficult conditions. This paper reports on the training of an object detection system using a range of traffic scenarios: sunny, rainy, snowy, one-side road, two-side road, complex road structures with occlusions, heavy traffic with congestion, light traffic, and reduced traffic at night. The state-of-the-art object detector of YOLOv5x is used for vehicle detection and is fine-tuned on this new diverse dataset through transfer learning. Transfer learning freezes the backbone network while training the remaining fully connected network. To further improve the detection performance, we added two convolutional block attention modules (CBAM) to the neck as our proposed system: 2xCBAM-YOLOv5. Several experiments refined the number of CBAMs and the placement of these modules to optimize performance. Doing transfer learning alone, the mean Average Precision(mAP) on the test data improves from 75.9% to 78.9%. After transfer learning, ablations were done on YOLOv5x combined with the new CBAMs. The resulting mAP reaches 85.0%, while precision improves from 82.3% to 88.2%, recall improves from 72.3% to 80.4% and F1-score improves from 0.77 to 0.841 compared with transfer learning alone. This new architecture provides an overall improvement for ITS traffic surveillance applications.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"14 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":"125427774","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}

{"title":"Adversarial Machine Learning Attacks in Internet of Things Systems","authors":"Rachida Kone, Otily Toutsop, K. Wandji, Kevin T. Kornegay, Joy Falaye","doi":"10.1109/AIPR57179.2022.10092216","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092216","url":null,"abstract":"Researchers are looking into solutions to support the enormous demand for wireless communication, which has been exponentially increasing along with the growth of technology. The sixth generation (6G) Network emerged as the leading solution for satisfying the requirements placed on the telecommunications system. 6G technology mainly depends on various machine learning and artificial intelligence techniques. The performance of these machine learning algorithms is high. Still, their security has been neglected for some reason, which leaves the door open to various vulnerabilities that attackers can exploit to compromise systems. Therefore, it is essential to evaluate the security of machine learning algorithms to prevent them from being spoofed by malicious hackers. Prior research has shown that the decision tree is one of the most popular algorithms used by 80% of researchers for classification problems. In this work, we collect the dataset from a laboratory testbed of over 100 Internet of things (IoT) devices. The devices include smart cameras, smart light bulbs, Alexa, and others. We evaluate classifiers using the original dataset during the experiment and record a 98% accuracy. We then use the label-flipping attack approach to poison our dataset and record the output. As a result, flipping 10%, 20%, 30%, 40%, and 50% of the poison data generated accuracies of 86%, 74%, 64%, 54%, and 50%, respectively.","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":"133823436","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}

{"title":"Exploring the Explainability of Machine Learning Algorithms for Prostate Cancer","authors":"D. Provenzano, Y. Rao, M. Loew, S. Haji-Momenian","doi":"10.1109/AIPR57179.2022.10092214","DOIUrl":"https://doi.org/10.1109/AIPR57179.2022.10092214","url":null,"abstract":"Convolutional Neural Networks (CNNs) have become revolutionary for identification of clinically significant (CS) prostate cancer (PCA) on prostate MRI. These algorithms suffer from an explainability problem however, and few have been implemented in clinical practice. We investigated explainability and various normalization techniques for PCA by optimizing the computational attributes that correspond to those that radiologists typically consider, including contrast and homogeneity. An open-source ProstateX dataset consisting of 330 (254 non-clinically significant, 76 clinically-significant cancerous) T2-weighted prostate MRIs and a corresponding previously trained high accuracy CNN (a ResNet) were used to evaluate causes of high performance. Data were preprocessed to isolate differences in contrast and homogeneity through: (1) varying the contrast to be 50%, 75%, 100%, 150%, and 175% of the original image by varying alpha in the OpenCV module in Python, (2) Thresholding the image (where pixels < 45%, 50%, and 55% of the average were set to 0), and (3) applying Canny edge detection. The ResNet model was retrained on each preprocessed dataset with the same set of parameters as was used for the initial model, and performance was evaluated. Model results were evaluated through 5-fold cross validation. Model performance at baseline on the MRI data matched the performance of the original winners of the ProstateX challenge (AUC 0.84). Thresholding the image at 50% (AUC 0.87) and increasing contrast of the image to 1.5x the baseline (AUC 0.83) also combined with Canny Edge Detection (AUC 0.83) matched baseline performance, indicating these features were important to the original model. Decreasing the contrast by 0.75x or 0.50x universally decreased model performance. Use of three popular normalization techniques to isolate features of interest and possible explainability in a machine learning model for prostate cancer found that no normalized data was able to outperform a baseline model. However, several perturbations resulted in statistically significantly decreased performance. These results indicated that if contrast is reduced from a baseline or if Canny edge detection is used without increased whole-image contrast, model performance suffers. Future research into normalization techniques is recommended to continue to improve prostate cancer model performance.","PeriodicalId":301162,"journal":{"name":"2022 IEEE Applied Imagery Pattern Recognition Workshop (AIPR)","volume":"49 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":"122683770","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}