train loss is to high when i run train.py in train/voc-fcn.is it normal?

[zz10001]

[zhangjiyuzjy]

modify loss function parameter

[zz10001]

modify loss function parameter

can you tell the detail,i have try many times,but it don't works.

[zhangjiyuzjy]

size_average set to True

[zz10001]

thx,i will try it

[zhangjiyuzjy]

do you test other models

[zhangjiyuzjy]

if you test and get good result ,can you give me some help

[zz10001]

if you test and get good result ,can you give me some help
sorry,i haven't try others model,because i just download the voc datasets.i just run voc-fcn.and the loss problem.i didn't run others.

[zhangjiyuzjy]

this mode can run but the prediction result is poor,maybe is my fault,if you get good result,please connect me

[zhangjiyuzjy]

i train this model loss is 0.07,but prediction is so poor

[zz10001]

this mode can run but the prediction result is poor,maybe is my fault,if you get good result,please connect me
it seem works when i modify size_average set to True.but the loss is still...like this.is it better after 100 epoch, if i get better result i will tell you!

[zhangjiyuzjy]

you use benchmark dataset,this need more time,thank you share go on to run