Researchers have developed a new model to understand how input resolution affects the latency of real-time object detection systems. The model treats total latency as a convolution of preprocessing, inference, and postprocessing stages, with parameters dependent on the source image resolution. Experiments using YOLOv11n on an NVIDIA Jetson Orin NX with COCO2017 images demonstrated that this resolution-aware parameterization can improve distributional approximations, particularly with more flexible models like Normal and Gamma. AI
IMPACT Provides a framework for optimizing real-time object detection systems by accounting for input resolution's effect on latency.
RANK_REASON Academic paper detailing a new model for object detection latency. [lever_c_demoted from research: ic=1 ai=1.0]
- Anderson-Darling Parameter Estimation of the Marshall-Olkin Length-Biased Exponential Distribution with Applications in Engineering and Environment
- arXiv
- COCO2017
- Cramer-von Mises and Anderson-Darling goodness of fit tests for extreme value distributions with unknown parameters
- Gamma
- Kolmogorov–Smirnov test
- NVIDIA Jetson Orin NX 16GB
- YOLOv11n
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
