ssd_mobilenet_v1_coco How is SSD implemented with mobilenet

Question

I'd like to understand how SSD has been trained with mobilenet. Has mobilenet been taken as a pretrained model through transfer learning, and then SSD has taken its weights to start another training?

Please guidance on this model.

Thanks.

Answer 1

MobileNet is the backbone of SSD in this case, or in other words, served as the feature extractor network. The original SSD was using VGG for this task, but later other variants of SSD started to use MobileNet, Inception, and Resnet to replace it.