soonminhwang/caffe-segmentation

Train/eval code for DilatedConv [Yu, ICLR'16], DeepLab-v2 [Chen, Arxiv'16], PSPNet [Zhao, CVPR'17]

caffe-segmentation's Issues

crop size problem

Hi, I use pspnet101_VOC2012_trian.prototxt add the data layer and loss layer ,when I set crop=347, the error is cudaSuccess out of memory, when I set crop=147, error is number of labels must match number of predictions, could you help me with that? BTW, i also confuse how to make sure source image size is same as label image size?

import util

i dont see the definition of "util"
someone succeed to run the predictiction?

thank

Training with pspnet

Hi @SoonminHwang How can i run training with PSPNet with https://github.com/hszhao/PSPNet/

any one train this caffe version in your own datasets? the accuracy is low in my datasets, i donot know why

any one train this caffe version in your own datasets? the accuracy is low in my datasets, i donot know why

Did you make it

Did you make it

The result of pspNet

I want to know what the result of pspnet is for this repo?

voc_eval.m error

>> VOCeval
Not enough input arguments.
Error in VOCevalseg (line 39)
    resfile = sprintf(VOCopts.seg.clsrespath,id,VOCopts.testset,imname);

Error in VOCeval (line 14)
[ acc, macc, conf, rawcounts ] = VOCevalseg(VOCopts);

voc_evalseg functionn require arguments id