I notice that the model which you train is using fssd frame,but when test time,you test your model by using ssd frame.The AddExtraLayers between ssd and fssd is different.Should I change score_ssd_pascal.py or ssd_detect.py to test my model?
By the way, I want to train fssd and using resnet replace vgg,which layers you think are the important layers to extract feature maps?
Usually, we can get the result of voc2007 in the training process.
If you want to replace vgg with resnet. I suggest fusing the last feature map from stage 3,4 and 5
I notice that the model which you train is using fssd frame,but when test time,you test your model by using ssd frame.The AddExtraLayers between ssd and fssd is different.Should I change score_ssd_pascal.py or ssd_detect.py to test my model?
By the way, I want to train fssd and using resnet replace vgg,which layers you think are the important layers to extract feature maps?
Thanks for you help