Xiaoyu SONG My personal Blog | Xiaoyu SONG

<< All Blogs

Week #28

The input T1w image tensor size: [1,128,128], the brain mask tensor size:[128,128].

Train with Resnet

For each group, with one subject, there are 3300 images for training. Batch size is set to 4. Initiate learning rate is 0.001, decrease by 0.85 every 100 steps. Parameter: 2,781,905

Here is the result of each group. Still running.s

group id iteration Loss batch size
0 999 0.0195 4
1 999 0.07927 4
2 999 0.1305 4
3 999 0.1350 4
4 999 0.12213 4
5 999 0.11824 4
6 999 0.122 4
7 999 0.124 4
8 999 0.12 4
9 999 0.1172 4
10 999 0.1157 4
11 999 0.1174 4
12 999 0.1211 4
13 999 0.1159 4
14 999 0.0701 4
15 999 0.0164 4
16 999 0.0014 4
17 999 0 4
18 999 0 4