This code line is here:
Firstly, the first scale feature map is drawn from
mobilenetv1 , however when input data is 300 * 300, this layer’s feature map is 19 * 19, or actually
18 * 18 using
model_zoo folder, which means that the step is actually 16.
Then, the step parameter in function
ssd_300_mobilenet0.25_coco is started from 8, and remains are
[8, 16, 32, 64, 100, 300] . I noted that the steps are used within anchor generation, and I want to know the step settings here could impact the training accuracy (mAP)?