Deeplab accuracy output
117 views
Skip to first unread message
韩靖
Dec 4, 2016, 9:04:07 PM12/4/16
to Caffe Users
I notice there are three accuracy output when deeplab train and test.
So, i want to know what does each of them represent.Thanks.
First is train
Second is test
So, i want to know what does each of them represent.Thanks.
First is train
Second is test
Eric J
Jan 20, 2017, 11:23:23 AM1/20/17
to Caffe Users
Hi, Jing,
Do you find the meaning of each accuracy? I also met the same question.
hmchan...@gmail.com
Jan 20, 2017, 5:26:18 PM1/20/17
to Caffe Users
As indicated in this file:
./src/caffe/layers/seg_accuracy_layer.cpp
in lines 102 to 104, you can see that
they correspond to accuracy, average recall, and average Jaccard. You can google confusion matrix for their definitions.
./src/caffe/layers/seg_accuracy_layer.cpp
in lines 102 to 104, you can see that
top[0]->mutable_cpu_data()[0] = (Dtype)confusion_matrix_.accuracy();
top[0]->mutable_cpu_data()[1] = (Dtype)confusion_matrix_.avgRecall(false);
top[0]->mutable_cpu_data()[2] = (Dtype)confusion_matrix_.avgJaccard();they correspond to accuracy, average recall, and average Jaccard. You can google confusion matrix for their definitions.
Reply all
Reply to author
Forward
0 new messages