AHOD: Adaptive Hybrid Object Detector for Context-Aware and real-time object detection in complex environments
Preprint
- 5 June 2025
- preprint
- Published by Springer Nature in Research Square
Abstract
This article introduces Adaptive Hybrid Object Detector (AHOD), a new paradigm in object detection designed to combine speed, accuracy and contextual adaptability. Current models, such as YOLO and Faster R-CNN, have significant limitations: YOLO excels in speed but often fails on complex objects, while Faster R-CNN prioritizes accuracy at the expense of inference time. AHOD is based on three major innovations: Feature Pyramid Enhancement (FPE), which improves multi-scale detection; Dynamic Context Module (DCM), which dynamically adjusts features according to context; and Fast and Accurate Detection Head (FADH), which balances speed and accuracy. Experimental results on COCO and Pascal VOC datasets show that AHOD outperforms existing models with an average accuracy (mAP) increase of 7%, while reducing inference time by 30%. These results demonstrate the potential of AHOD for real-time critical applications such as autonomous vehicles and surveillance.Keywords
All Related Versions
This publication has 14 references indexed in Scilit:
- Adaptive Perception for Unified Visual Multimodal Object TrackingIEEE Transactions on Artificial Intelligence, 2025
- Design of an Integrated Model for Video Summarization Using Multimodal Fusion and YOLO for Crime Scene AnalysisIEEE Access, 2025
- Context-Driven Automatic Target Detection With Cross-Modality Real-Synthetic Image MergingIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025
- SCGC-Net: Spatial Context-Guided Calibration Network for Multisource RSI Landslides DetectionIEEE Transactions on Geoscience and Remote Sensing, 2025
- Improving UAV Aerial Imagery Detection Method via Superresolution SynergyIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2025
- Pest-PVT: A model for multi-class and dense pest detection and counting in field-scale environmentsComputers and Electronics in Agriculture, 2024
- Vision language model for interpretable and fine-grained detection of safety compliance in diverse workplacesExpert Systems with Applications, 2024
- Towards improving context-aware applications using link quality induction and self-calibrated ultra-wide band localization systemsComputing, 2024
- A systematic literature review of recent advances on context-aware recommender systemsArtificial Intelligence Review, 2024
- Fast video anomaly detection via context-aware shortcut exploration and abnormal feature distance learningPattern Recognition, 2024