search

alexgkendall / SegNet-Tutorial

Files for a tutorial to train SegNet for road scenes using the CamVid dataset
http://mi.eng.cam.ac.uk/projects/segnet/tutorial.html
847 stars 517 forks source link

File "Scripts/test_segmentation_camvid.py", line 35, in <module> label = net.blobs['label'].data KeyError: 'label' #82

Open yonglinZ opened 7 years ago

yonglinZ commented 7 years ago

I0226 17:20:08.784281 16551 net.cpp:194] relu1_2 does not need backward computation. I0226 17:20:08.784282 16551 net.cpp:194] conv1_2_bn does not need backward computation. I0226 17:20:08.784284 16551 net.cpp:194] conv1_2 does not need backward computation. I0226 17:20:08.784306 16551 net.cpp:194] relu1_1 does not need backward computation. I0226 17:20:08.784307 16551 net.cpp:194] conv1_1_bn does not need backward computation. I0226 17:20:08.784309 16551 net.cpp:194] conv1_1 does not need backward computation. I0226 17:20:08.784312 16551 net.cpp:235] This network produces output argmax I0226 17:20:08.784343 16551 net.cpp:482] Collecting Learning Rate and Weight Decay. I0226 17:20:08.784406 16551 net.cpp:247] Network initialization done. I0226 17:20:08.784409 16551 net.cpp:248] Memory required for data: 1065139200 Traceback (most recent call last): File "Scripts/test_segmentation_camvid.py", line 35, in label = net.blobs['label'].data KeyError: 'label'

In " test_segmentation_camvid.py" , I only modified caffe.set_mode_cpu() and the caffe_root. Please help me solve this problem. THX.

yonglinZ commented 7 years ago

I have tried all of the caffemodels from SegNet Model Zoo, but still the same problem KeyError: 'label'. Anyone can help? Here is the cammond: python Scripts/test_segmentation_camvid.py --model Example_Models/segnet_model_driving_webdemo.prototxt --weights Example_Models/segnet_basic_camvid.caffemodel --iter 233

jjdblast commented 7 years ago

I'm using gpu docker and got this issue too

yonglinZ commented 7 years ago

Maybe, I think the caffemodel have some problem. I am trying to retrain the caffemodel according to the tutorial.

jjdblast commented 7 years ago

@yonglinZ Cool then, how long it'll take using caffe to train? I train a basci version of segnet using tensorlfow on K80, which tooks me about 8 hours, but it can't performance as well as the web demo.

smohiudd commented 6 years ago

I'm running into the same problem. Has anyone been able to figure this out?

labibi commented 6 years ago

@smohiudd @yonglinZ did you solve it?