prepare my own data set and creat LMDB format data set

359 views
Skip to first unread message

zofnat bar

unread,
Jul 5, 2016, 2:31:50 AM7/5/16
to Caffe Users
Hello,

I'm new to Caffe and I'm willing to know how to prepare my own data set:  (i want to creat LMDB format data set)

I'm understand that i need to prepare:

1 folder train: the inside of the training picture  --- > in this folder i put my pictures to train- jpg file?

2 Folder Val: which put the picture of Val --- > but whats is mean picture of val ? 

3 train.txt: the file name of the training picture and the corresponding category  ----> where i put this file? and how is look like?

4 val.txt: test images of the file name and the corresponding category --- > where i put this file? and how is look like?

 

in yellow is the questions.


Thanks :)
zofnat

Ahmed Ibrahim

unread,
Jul 5, 2016, 1:15:37 PM7/5/16
to Caffe Users
check the script caffe/examples/imagenet/create_imagenet.sh and you will learn how to do it easily

zofnat bar

unread,
Jul 12, 2016, 1:03:19 AM7/12/16
to Caffe Users
hi again , thanks a lot.
now after i training my data set , the model based on about 239 images which is divided in to 3 categories (labels).
when i try to classify image i have shape mismatch.

I run this:
~/Documents/caffe$ ./build/examples/cpp_
classification/classification.bin   models/bvlc_reference_caffenet/deploy.prototxt   models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel   data/ilsvrc12/imagenet_mean.binaryproto   data/ilsvrc12/synset_words.txt   examples/images/cat.jpg



F0711 16:44:09.304422  5779 net.cpp:765] Cannot copy param 0 weights from layer 'fc6'; shape mismatch.  Source param shape is 1 1 4096 9216 (37748736); target param shape is 3 9216 (27648). To learn this layer's parameters from scratch rather than copying from a saved net, rename the layer.
*** Check failure stack trace: ***
    @     0x7fd5aa7f8daa  (unknown)
    @     0x7fd5aa7f8ce4  (unknown)
    @     0x7fd5aa7f86e6  (unknown)
    @     0x7fd5aa7fb687  (unknown)
    @     0x7fd5aac009a7  caffe::Net<>::CopyTrainedLayersFrom()
    @     0x7fd5aac08262  caffe::Net<>::CopyTrainedLayersFromBinaryProto()
    @     0x7fd5aac082c6  caffe::Net<>::CopyTrainedLayersFrom()
    @           0x407505  Classifier::Classifier()
    @           0x403feb  main
    @     0x7fd5a71a6f45  (unknown)
    @           0x4044fb  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)

thanks in advance :)


בתאריך יום שלישי, 5 ביולי 2016 בשעה 20:15:37 UTC+3, מאת Ahmed Ibrahim:
Reply all
Reply to author
Forward
Message has been deleted
0 new messages