NanoDet-Plus is a one-stage anchor-free object detection model based on FCOS architecture, using Generalized Focal Loss for classification and regression. It introduces a novel label assignment strategy with an assign guidance module and dynamic soft label assigner. The model achieves 34.3 mAP while maintaining real-time performance on mobile devices, with support for multiple inference backends including ncnn, MNN, and OpenVINO, plus Android deployment examples.