extract_features with mnist_autoencoder.prototxt

108 views
Skip to first unread message

Yingjun Li

unread,
Nov 5, 2015, 5:00:20 AM11/5/15
to Caffe Users

I want to extract the activation of encode4 layer. However, I get a 'Aborted' error. Can you please give me an idea about where the problem is?


Please see the below:
yingjun@OZ:~/caffe$ ./build/tools/extract_features ~/caffe/examples/mnist/mnist_autoencoder_iter_65000.caffemodel ~/caffe/examples/mnist/mnist_autoencoder.prototxt encode4 ~/test_features 10 lmdb GPU
E1105 17:17:23.871481 31882 extract_features.cpp:54] Using GPU
E1105 17:17:23.872267 31882 extract_features.cpp:60] Using Device_id=0
F1105 17:17:24.187971 31882 insert_splits.cpp:35] Unknown bottom blob 'data' (layer 'flatdata', bottom index 0)
*** Check failure stack trace: ***
@ 0x7f8fda70fdaa (unknown)
@ 0x7f8fda70fce4 (unknown)
@ 0x7f8fda70f6e6 (unknown)
@ 0x7f8fda712687 (unknown)
@ 0x7f8fdaa9d206 caffe::InsertSplits()
@ 0x7f8fdabf47fd caffe::Net<>::Init()
@ 0x7f8fdabf7578 caffe::Net<>::Net()
@ 0x40878a feature_extraction_pipeline<>()
@ 0x7f8fd9b3eec5 (unknown)
@ 0x4044de (unknown)
@ (nil) (unknown)
Aborted (core dumped)

Jan C Peters

unread,
Nov 6, 2015, 3:08:22 AM11/6/15
to Caffe Users
For this you need to use a 'deploy' version of the network prototxt. For further info look at my following answer in this thread: https://groups.google.com/d/msg/caffe-users/zBDa8kSlRR0/gzVUZjcBDAAJ

Jan
Reply all
Reply to author
Forward
0 new messages