Why FCN32 trained on VOC2011 training data has different results in the paper

[Sharif Amit]

Hello
As u can see in the paper Fully Convolutional Neural Networks
the test results  on page 5, in the table 1 states that the FCN_VGG16 (which is the FCN 32) trained on only VOC2011 training data has a mIOU of 56.0( for a fixed learning rate). But in page 7, Section 4.4 (More training data) it states that they have increased its validation score from 57.7 (trained on only VOC2011 data) to 63.6 percent with additional SBDdataset. So my question is, did they change the learning parameter to get the latter one which is 57.7 percent  ?

[Evan Shelhamer]

Table 1 compares different FCN architectures with the optimization settings described in section 4.1. As explained by section 4.2 online, high momentum optimization does better. 57.7 is the mean IU obtained by learning FCN-32s on VOC11 segtrain alone.

Evan Shelhamer

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users+unsubscribe@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/4c19b4e2-6a90-47ff-9f53-e548fc9de67c%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Sharif Amit]

Thank you Evan.