segmentation fault error in training caffe based convolutional neural network model

[gaurav pawar]

I am following a tutorial for human segmentation, in which a FCN-8s caffe model is trained for semantic segmentation problem using Pascal VOC-12 dataset in which out of 21 classes, one class is of person, and on top of it they have again trained the similar model but for 1 class, which is human in portrait mode, and images in training database is of 800 X 600 size. But when i tried it to train for landscape mode images of human of size 360 X 640, i am getting segmentation fault error while setting up fc6 convolutional layer, error log is in the below link as its too big to post it here. error log

training script

protobuf file for FCN-8s model

solver file for model used on top of FCN-8s

train protobuf file for model used on top of FCN-8

[Angeline Y.]

Hi Gaurav,

I'm facing the same problem. Did you manage to find a solution?