Yalong Pi

Yalong Pi

orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 January 2022 in Computational Urban Science
Verified
Journal Article
Open access
PDF file
YPYalong Pi
Yalong Pi
NDNick Duffield
Nick Duffield
ABAmir H. Behzadan
Amir H. Behzadan
TLTim Lomax
Tim Lomax
Accurate and prompt traffic data are necessary for the successful management of major events. Computer vision techniques, such as convolutional neural network (CNN) applied on video monitoring data, can provide a cost-efficient and timely alternative to traditional data collection and analysis methods. This paper presents a framework designed to take videos as input and output traffic volume counts and intersection turning patterns. This framework comprises a CNN model and an object tracking algorithm to detect and track vehicles in the camera’s pixel view first. Homographic projection then maps vehicle spatial-temporal information (including unique ID, location, and timestamp) onto an orthogonal real-scale map, from which the traffic counts and turns are computed. Several video data are manually labeled and compared with the framework output. The following results show a robust traffic volume count accuracy up to 96.91%. Moreover, this work investigates the performance influencing factors including lighting condition (over a 24-h-period), pixel size, and camera angle. Based on the analysis, it is suggested to place cameras such that detection pixel size is above 2343 and the view angle is below 22°, for more accurate counts. Next, previous and current traffic reports after Texas A&M home football games are compared with the framework output. Results suggest that the proposed framework is able to reproduce traffic volume change trends for different traffic directions. Lastly, this work also contributes a new intersection turning pattern, i.e., counts for each ingress-egress edge pair, with its optimization technique which result in an accuracy between 43% and 72%.

13 September 2022 in Urban Science
Verified
Journal Article
Open access
PDF file
YPYalong Pi
Yalong Pi
Xinyue YeXinyue Ye
Xinyue Ye
NDNicholas Duffield
Nicholas Duffield
OGon behalf of the Microsoft AI for Humanitarian Action Group
on behalf of the Microsoft AI for Humanitarian Action Group
The winter storm Uri that occurred in February 2021 affected many regions in Canada, the United States, and Mexico. The State of Texas was severely impacted due to the failure in the electricity supply infrastructure compounded by its limited connectivity to other grid systems in the United States. The georeferenced estimation of the storm’s impact is crucial for response and recovery. However, such information was not available until several months afterward, mainly due to the time-consuming and costly assessment processes. The latency to provide timely information particularly impacted people in the economically disadvantaged communities, who lack resources to ameliorate the impact of the storm. This work explores the potential for disaster impact estimation based on the analysis of instant social media content, which can provide actionable information to assist first responders, volunteers, governments, and the general public. In our prototype, a deep neural network (DNN) uses geolocated social media content (texts, images, and videos) to provide monetary assessments of the damage at zip code level caused by Uri, achieving up to 70% accuracy. In addition, the performance analysis across geographical regions shows that the fully trained model is able to estimate the damage for economically disadvantaged regions, such as West Texas. Our methods have the potential to promote social equity by guiding the deployment or recovery resources to the regions where it is needed based on damage assessment.

8 February 2023 in Traffic Injury Prevention
Verified
Journal Article
Closed access
YPYalong Pi
Yalong Pi
NDNick Duffield
Nick Duffield
ABAmir Behzadan
Amir Behzadan
TLTim Lomax
Tim Lomax
Work zone speed is one of the most important factors in road construction safety management. This work presents a computer vision based technique designed to measure lane-specific individual vehicl...

18 August 2020 in CIB W78 Proceedings
Verified
Conference Paper
Open access
YPYalong Pi
Yalong Pi
NNNipun D Nath
Nipun D Nath
ABAmir H Bezhadan
Amir H Bezhadan
The ITC Digital Library has an ambition to provide a single point of entry to scholarly and research publication from the domain of construction informatics alias construction information technology. We believe that works, in full text, should be available for free, to the researchers, students and the industry.

9 March 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
YPYalong Pi
Yalong Pi
UVUday Vysyaraju
Uday Vysyaraju
APAkash Pillai
Akash Pillai
SLShoujia Li
Shoujia Li
JWJack Wooley
Jack Wooley
NDNicholas Duffield
Nicholas Duffield
Accurate and timely preliminary damage assessment (PDA) following natural disasters is critical for effective emergency response and resource allocation. Volunteered geographic information (VGI) such as social media posts including text, image, and others provide impact signals which can be rapidly processed by customized neural networks. This study introduces a disaster impact estimation prediction (DIEP) that takes social media self-reports and outputs impact probabilities for FEMA defined classes (following names use exact OpenFEMA names) including homeOwnersInsurance, floodInsurance, destroyed, floodDamage, roofDamage, tsaEligible (temporary shelter eligible), tsaCheckedIn, rentalAssistanceEligible, repairAssistanceEligible, replacementAssistanceEligible, and personalPropertyEligible. The DIEP model is a natural language processing (NLP) model that is trained and tested target data surveyed and published by OpenFEMA at zip code level. Due to the social media (Twitter in this study) data availability, the focus of this work is on tropical storms in Texas from 2010 to 2024: Hurricanes Harvey, Hanna, Laura, Imelda, and Beryl. Around a million tweets with geo-tags, 2.2 million FEMA records at household level are collected, organized, and published. The BERT based DIEP outperformed others, achieving up to 98% precision and recall in several categories, when classifying zip code level impact. Cross-storm validation demonstrates the model’s generalizability, with strong F1 score performance in predicting homeowner insurance (98.72%), flood insurance (81.01%), flood damage (86.57%), and temporary shelter eligible (92.52%), especially when trained on historical events (Harvey and Imelda) and predict future events (Beryl). The DIEP system also supports real-time visualization of disaster progression, offering a scalable and cost-effective tool for first responders and policymakers.

8 January 2026 in Journal of Animal Science
Verified
Journal Article
Open access
YPYalong Pi
Yalong Pi
Karun KaniyamattamKarun Kaniyamattam
Karun Kaniyamattam
AAAdeolu Adekunle
Adeolu Adekunle
AAAniruddh Adiga
Aniruddh Adiga
LTLuis Tedeschi
Luis Tedeschi
CZCheng Zhang
Cheng Zhang
Precision livestock farming (PLF) increasingly relies on computer vision (CV) to monitor animal behavior, health, and production. However, widespread adoption remains limited by the large quantities of manually labeled data required to train most artificial intelligence (AI) models. This paper reviews AI and CV applications in PLF and presents three data-efficient case studies conducted at a research feedlot at Texas A&M University. The assessment of AI model predictability was conducted with precision, recall, F1-score, and adjusted rand index (ARI). In Case Study 1, a data-efficient Mask R-CNN (region-based convolutional neural network) model pre-trained on the common objects in context (COCO) dataset was applied without modification to detect cattle feeding events at individual feed bunks. Across 98,720 frame-bunk classifications, the system achieved 87.2% precision, 89.1% recall, and an 88.2% F1 score compared to physical bunk sensor ground truth. In Case Study 2, transfer learning on a ResNet backbone was used to classify bunk feed grade across six customized levels (S00–S40) using only 183 labeled images. The model achieved 91% accuracy on an independent verification set of 7,905 images. In Case Study 3, statistical tools, i.e. Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and k-means clustering, were applied to identify individual animals at the bunk without manual animal-level annotation. This method and achieved an ARI of 0.96 against expert-verified ground truth, successfully differentiating animals. Integrating all three components, the framework produced a continuous record of individual animal bunk visits and estimated feed intake from a 1-h video. Key remaining challenges include long-term individual animal re-identification under changing appearance and environmental conditions, incorporation of spatiotemporal information to better characterize feeding behavior, and the limited availability of large-scale open datasets for PLF research. Future research will integrate multispectral imaging and audio analytics to improve model robustness and advance individual animal identification for cattle health and nutritional management.