questions about pretrain model vgg16_caffe-292e1171.pth

[lyxlynn]

What is the difference between vgg16_caffe-292e1171.pth and ssd300_vgg16_coco-b556d3b4.pth?
Is vgg16_caffe-292e1171.pth trained from imagenet ?
Is ssd300_vgg16_coco-b556d3b4.pth trained from coco dataset?

I use my own datas to train ssd300 , the results are strange.
use vgg16_caffe-292e1171.pth (pretrain): mAP 70.0
use ssd300_vgg16_coco-b556d3b4.pth ( pretrain) mAP:32
use ssd300_vgg16_coco-b556d3b4.pth (load from )

pretrain: only use backbone
load from: full model including neck and head