2023 IEEE Conference on Artificial Intelligence (CAI)最新文献

{"title":"Automated Fracture Detection from CT Scans","authors":"Abhishek Kumar Chaudhary, Shotabdi Roy, Rodrigue Rizk, K. Santosh","doi":"10.1109/CAI54212.2023.00077","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00077","url":null,"abstract":"Computed Tomography (CT) scans play a crucial role in modern medical imaging for detecting bone fractures. However, identifying the location and position of broken bones can be challenging, particularly in complex cases involving multiple extremities. In this paper, we propose a robust approach for enhancing fracture detection and localization in CT scans using the YOLO v7 model. By simultaneously predicting class probabilities and bounding boxes in a single iteration, the YOLO v7 model shows improved and consistent performance measures. We developed our approach on a dataset of 1217 CT cases, by training our model on combined extremities, resulting in improved and consistent performance metrics for detecting and localizing fractures. Our proposed method achieved a high precision rate of 99% for identifying broken bones in the lower right limb and 66% for the combined set of upper and lower extremities on both sides. Our findings highlight the potential of YOLO v7 as a powerful tool for enhancing medical imaging workflows, particularly for further treatment planning, by improving fracture detection and localization. Future studies could investigate the generalizability and scalability of our proposed method in larger datasets and different clinical settings.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115373900","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":"Safety Margins for Reinforcement Learning","authors":"Alexander Grushin, Walt Woods, Alvaro Velasquez, Simon Khan","doi":"10.1109/CAI54212.2023.00026","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00026","url":null,"abstract":"Any autonomous controller will be unsafe in some situations. The ability to quantitatively identify when these unsafe situations are about to occur is crucial for drawing timely human oversight in, e.g., freight transportation applications. In this work, we demonstrate that the true criticality of an agent’s situation can be robustly defined as the mean reduction in reward given some number of random actions. Proxy criticality metrics that are computable in real-time (i.e., without actually simulating the effects of random actions) can be compared to the true criticality, and we show how to leverage these proxy metrics to generate safety margins, which directly tie the consequences of potentially incorrect actions to an anticipated loss in overall performance. We evaluate our approach on learned policies from APE-X and A3C within an Atari environment, and demonstrate how safety margins decrease as agents approach failure states. The integration of safety margins into programs for monitoring deployed agents allows for the real-time identification of potentially catastrophic situations.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126172588","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":"Automated energy management and learning","authors":"Gabriel Santos, B. Teixeira, T. Pinto, Z. Vale","doi":"10.1109/cai54212.2023.00037","DOIUrl":"https://doi.org/10.1109/cai54212.2023.00037","url":null,"abstract":"Automatic energy management systems allow users’ active participation in flexibility management while assuring their energy demands. We propose a transparent framework for automated energy management to increase trust and improve the learning process, combining machine learning, experts’ knowledge, and semantic reasoning. A practical example of thermal comfort shows the advantages of the framework.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115253191","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":"Exploration and Comparison of Locomotion Mode Recognition Models for Prosthetic Gait","authors":"A. Gouda, J. Andrysek","doi":"10.1109/cai54212.2023.00072","DOIUrl":"https://doi.org/10.1109/cai54212.2023.00072","url":null,"abstract":"Establishing generalizable models for locomotion mode recognition (LMR) of prosthetic gait can be challenging due to limited access of sufficient labelled datasets. Hence, subject-specific models continue to be primarily used. However, there are no studies that investigated the effect of reducing the amount of training data that is presented to the machine learning model during training. Additionally, previously validated LMR models for prosthetic gait primarily used LDA classifiers. However, literature suggests that RF models may improve overall accuracy based on able-body validation. Therefore, to address those gaps, this study compared the performance of LDA and RF models for prosthetic gait and classifiers to LDA. Varied test size ratios data were evaluated to assess the trade-off between performance and amounts of training data.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"169 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114252326","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":"Incorporating Temporal and Meteorological Data for Generating Pseudo-measurements in Active Distribution Power Networks","authors":"Sepideh Radhoush, Kaveen Liyanage, Trevor C. Vannoy, Bradley M. Whitaker, H. Nehrir","doi":"10.1109/CAI54212.2023.00027","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00027","url":null,"abstract":"This paper proposes a new data-based algorithm to generate pseudo-measurements in active distribution networks with a high penetration of distributed generations and a limited number of real measurements. The real measurements, along with temporal and meteoritical features, are considered in order to improve the performance of the pseudo-measurement results. The proposed method consists of a set of base learners and a meta learner to generate pseudo-measurements. The proposed method is compared against a model with training data divided based on a random split without consideration of time, temperature, and seasonality data. To ensure the model training is robust against different dynamic behavior, temporal and meteorological information from Bozeman, MT, USA are considered. Furthermore, pseudo-measurements generated using the proposed method, along with the real measurements, are fed into a Weighted Least Squares method to perform state estimation calculations. The effectiveness of our proposed method is evaluated using a modified IEEE standard 69 bus distribution.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130381360","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":"Automated pallet handling via occlusion-robust recognition learned from synthetic data*","authors":"Csaba Beleznai, Marcel Zeilinger, Johannes Huemer, Wolfgang Pointner, Sebastian Wimmer, P. Zips","doi":"10.1109/CAI54212.2023.00039","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00039","url":null,"abstract":"Vision-based perception is a key enabling technology when attempting to convert human work processes into automated robotic workflows in diverse production and transport scenarios. Automation of such workflows, however, faces several challenges due to the diversity governing these scenarios: various objects to be handled, differing viewing conditions, partial visibility and occlusions. In this paper we describe the concept of an occlusion-robust pallet recognition methodology trained fully in the synthetic domain and well coping with varying object appearance. A key factor in our representation learning scheme is to entirely focus on geometric traits, captured by the surface normals of dense stereo depth data. Furthermore, we adopt a local key-point detection scheme with regressed attributes allowing for a bottom-up voting step for object candidates. The proposed geometric focus combined with local key-point based reasoning yields an appearance-independent (color, texture, material, illumination) and occlusion-robust detection scheme. A quantitative evaluation of recognition accuracy for two network architectures is performed using a manually fine-annotated multi-warehouse dataset. Given the standardized pallet dimensions, spatially accurate pose estimation and tracking, and robotic path planning are carried out and demonstrated in two automated forklift demonstrators. These demonstrators exhibit the ability to consistently perform automated pick-up and drop-off of pallets carrying arbitrary items, under a wide variation of settings.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130630558","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":"Electricity Consumption Prediction via WaveNet+t","authors":"Xiuxuan Sun, Jianhua Chen","doi":"10.1109/cai54212.2023.00033","DOIUrl":"https://doi.org/10.1109/cai54212.2023.00033","url":null,"abstract":"Electricity consumption prediction is essential for load management to prevent shortage and excess supply. Different methods ranging from statistical methods, machine learning, and deep learning models were developed to predict electricity consumption. In this study, a probabilistic model -WaveNet+t was developed to provide the confidence interval rather than the deterministic estimate. WaveNet+t model integrates dilated causal convolutional neural networks with residual networks to extract the temporal, long/short term patterns from the time series data. The testing results based on a real dataset from 370 clients showed that WaveNet+t model has a lower CRPSs1״״ value than the benchmark models.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128015825","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":"Improving Idiopathic Pulmonary Fibrosis Damage Prediction with Segmented Images in a Deep Learning Model","authors":"Sheila Leyva-López, Gerardo Hernández-Nava, Enrique Mena-Camilo, Sebastián Salazar-Colores","doi":"10.1109/CAI54212.2023.00078","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00078","url":null,"abstract":"This work introduces a semantic segmentation model, UNet, as a preprocessing module to an algorithm predicting lung damage caused by Idiopathic Pulmonary Fibrosis. By modifying the model input through the incorporation of a guide image (a segmentation result) into the original image, we observed an improved performance of eight out of twelve tested backbones in the prediction model, with an improvement of up to 0.57 in the LLLm metric. This study underscores the significance of data preprocessing in deep learning models’ performance. The inclusion of additional data, such as segmented images, can significantly enhance a model’s ability to perform specific tasks, emphasizing the need for careful data preprocessing to obtain precise and reliable results when implementing deep learning models for lung damage prediction.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127553721","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":"Jet Engine Modulation Recognition with Deep Learning","authors":"Anne L. Lee, Phillip Ly, Marsh Jackson, E. Saint-Pierre, Phil Phuoc T. Ho, David Wilson","doi":"10.1109/CAI54212.2023.00011","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00011","url":null,"abstract":"Jet Engine Modulation (JEM) spectral characteristics can be used with deep learning neural networks to enhance Automatic Target Recognition (ATR) capability. The conventional approaches for ATR using JEM lines include the estimations of the blade count, frequency of rotation, symmetry of the spectral lines, delta spectrum from multiple compression stages, and other special features. These JEM features are being compared with baseline features stored in a database using nearest neighbor classification for a best match. The existing feature extraction logics are data driven and tuned to a limited data set. Therefore, we developed a JEM ATR with deep learning algorithm to identify the signal scattered returns from the engine structure in periodic modulation. The JEM ATR deep learning algorithm enables the optimization of rotating blades in jet engines modulation pattern by self-training through reinforcement learning as this is an incredible breakthrough for artificial intelligence. The JEM models include four high-fidelity targets with thirty epochs of deep learning optimizer runs. Initially, our JEM target recognition results in a confusion matrix to validate Model_A determined that Target 1, Target 2, and Target 3 are 100% primary targets over 400 training samples. Target 4 has a 21.3% chance of being false target over 400 training samples for Model_A. Subsequently, when the optimizer hyperparameters and other parameters are fine-tuned with more training and sampling sessions, the ATR accuracy increased to 100% for all four targets with Model_P. Our proposed method can drastically improve the accuracy of automatic target recognition capability for radar systems using JEM deep learning algorithms.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128935458","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":"Identifying Important Leisure-time Living Activities for Healthy Aging in the Singapore Longitudinal Aging Cohort Using Machine Learning Techniques","authors":"Wangyang Hu, Xin Zhong, Feng Yang","doi":"10.1109/CAI54212.2023.00144","DOIUrl":"https://doi.org/10.1109/CAI54212.2023.00144","url":null,"abstract":"Singapore's aging population has led to a government commitment to promoting healthy aging through the construction of smart and resilient communities. However, the design of effective community services can be challenging due to a lack of understanding of important leisure-time daily living activities that promote healthy aging. To address this issue, we developed a novel learning-based computational workflow to identify important living activities correlated with both clinical and biological health for healthy aging. Our analysis of 1356 community-living Chinese elderly in the Singapore Longitudinal Aging Study (SLAS) II cohort revealed that 10 living activities were significantly associated with clinically healthy aging, while 9 were significantly associated with biologically healthy aging through the selection of minimum number of features by 7 algorithms (Decision Tree, Random Forest, Naïve Bayes, Logistic Regression, K-Nearest Neighbor, Multilayer Perceptron and XGBoost). We compared two learning-based feature selection methods algorithms, Recursive Feature Elimination (RFE) and Sequential Forward Selection (SFS), and found that features selected by SFS method outperformed those by RFE method. Physical exercise and senior club activities were found to be important leisure-time daily-living activities. Further analysis indicated that the active group, composed of older adults who participated in these activities, had significantly longer survival times, a lower mortality rate (lifespan) and a lower frailty rate (healthspan) compared to the non-active group (p<0.001). The percentage of dead/frail people in the non-active group tripled. These findings demonstrate the potential impact of using machine learning techniques to assist healthy aging studies. This work links biological health (aging markers and biological age), clinical health and leisure-time daily living activities in SLAS cohort studies. By identifying and prioritizing these activities, policymakers and service providers can develop interventions that are evidence-based and culturally appropriate, maximizing their potential impact on the health and well-being of older adults in Singapore.","PeriodicalId":129324,"journal":{"name":"2023 IEEE Conference on Artificial Intelligence (CAI)","volume":"552 ","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133523461","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}