MobileNet-SSD

MobileNet-SSD is a lightweight, single-shot object detection network engineered for high-speed, real-time performance on resource-constrained mobile and embedded devices.

This is a high-efficiency object detection model, pairing the MobileNet backbone with the Single Shot MultiBox Detector (SSD) framework. The MobileNet component utilizes depthwise separable convolutions, drastically reducing computational cost and parameter count for deployment on edge devices. SSD enables a single forward pass to predict bounding boxes and class probabilities, eliminating slow region proposal steps. This architecture delivers fast inference speeds (high FPS) and a low memory footprint; one Caffe implementation achieved a 0.727 mAP on the VOC0712 dataset, confirming its reliable accuracy for real-time applications like security systems and augmented reality.

https://github.com/chuanqi305/MobileNet-SSD