How to run FCN for training and testing?

606 views
Skip to first unread message

john1...@gmail.com

unread,
Dec 14, 2016, 2:14:01 AM12/14/16
to Caffe Users
Hello caffe users,

I am new memember of caffe. I want to run the Fully convolution network for semantic segmentation. I downloaded and and build the caffe source. After that, I downloaded the FCN code at https://github.com/shelhamer/fcn.berkeleyvision.org

Then I download the VOC2011 data and put it in the data/pascal folder. The pascal folder contains

classes.txt  README.md  seg11valid.txt  VOC2011


Finally, I went to the folder voc-fcn8s which contains

caffemodel-url  deploy.prototxt  net.py  solve.py  solver.prototxt  train.prototxt  val.prototxt

First, I want to train the VOC2011 data and then I want to perform semantic segmentation. I could not find any step by step to run these python codes. If you was successful to train and test the data and FCN. Could you guide me how can I run them? Thank you in advance

souZou

unread,
Dec 14, 2016, 11:05:19 AM12/14/16
to Caffe Users
Hello,
I started to use FCN recently, and I want to run the training with nyud dataset. I downloaded FCN in the caffe-master source.
When start training i have this error

I1214 12:25:41.467631 29687 layer_factory.hpp:77] Creating layer color
ImportError: No module named nyud_layers


Thanks for help

Sara Caffe

unread,
Dec 18, 2016, 10:11:21 AM12/18/16
to Caffe Users
Hi John,

I want to run this code, were you successful to run it? if yes, can you please let me know what did you do exactly? 

Thanks
Sara

Arghavan Arafati

unread,
Jan 17, 2017, 1:55:10 PM1/17/17
to Caffe Users
Hello,

Have you solved the import error? I have the same problem. :/ I've check PYTHONPATH an re make everything but still have problem with pylayers.

Thanks,

Ara

Sreenu

unread,
Sep 27, 2017, 3:26:42 PM9/27/17
to Caffe Users
Hello john1

Did you run the FCN code, can you help me on this how to run this code.

Bandhanpreet Kaur

unread,
Apr 17, 2019, 2:38:59 PM4/17/19
to Caffe Users
Hey Sreenu, were you able to figure it out?
Reply all
Reply to author
Forward
0 new messages