Minimum Detectable Object Area in YOLOv11: Effects of Resolution and Model Scale

Detecting objects that occupy only a small fraction of an image remains a fundamental limitation of modern object detection systems. Although recent YOLO architectures employ multi-scale feature fusion to improve robustness across object sizes, the minimum object area required for reliable detection is rarely quantified explicitly. This paper presents a systematic, variant wise analysis of the detection limits of YOLOv11, focusing on how input image resolution, object scale, and model capacity jointly influence detectability. Experiments evaluate five object categories across image resolutions ranging from 2560×2560 to 40×40 pixels, progressively reducing object size while maintaining consistent evaluation conditions. The minimum detectable object area is defined as the smallest object consistently detected above a 70% confidence threshold and is reported as a percentage of total image area. Results show that input resolution strongly dominates detection feasibility, with minimum detectable area increasing sharply as resolution decreases. While larger YOLOv11 variants consistently outperform smaller models at moderate resolutions, all variants exhibit a resolution-dependent detection floor below which detection collapses regardless of model capacity. Additionally, object characteristics such as shape and texture significantly influence detection thresholds. These findings provide practical insights into the operational limits of YOLOv11 and inform model selection and resolution design for small object detection applications.