📄 Abstract
A new benchmark for the efficiency of neural networks was set in January 2026 when the YOLO26 architecture was introduced. This research assesses the performance of the YOLO26n (Nano) model in a real-time environment using a subset of the COCO dataset. The model reached a peak mAP@50 of 0.5122, validating its suitability for consumer-grade hardware. 🚀 Features
Batch Analytics: Upload image/video files for automated object density counts.
Local Edge Mode: Real-time interface with local hardware (webcams).
Optimized Performance: Inference speed of <12ms per frame on standard CPUs.
📊 Results
Accuracy: Peak mAP@50 of 0.5122.
Inference Speed: Support for smooth 30fps video streams.
📚 Citation
If you use this work in your research, please cite it as:
A, S. (2026). Real-Time Multi-Class Object Detection and Analytics using the YOLO26n Architecture (1.0). Zenodo. https://doi.org/10.5281/zenodo.18906733