trian/test.prototxt

[foralliance]

@Robert-JunWang

I found that in the final prediction, stage4_tb/ext/pm2/res was used to make two predictions, except that the min_size and max_size settings in the "PriorBox" were different, all the others are same

Why do you make two predictions separately instead of merging them? What are the advantages of these two predictions?

[Robert-JunWang]

You can merge them, the result should be the same. The reason why I use two predictions separately is to make things easier to try different feature extraction networks. Also, It makes possible to try the unified prediction mentioned in RUN with the last 5 feature maps, but I do not see any improvement on mAP by using unified prediction.

[foralliance]

many many thanks