creating lmdb

[Jalen Hawkins]

I am new to caffe and trying to learn how to create my own data base with some test images i downloaded. After finally learning how to i have ran into an error. It is showing that i have 0 files inside the data base.

./examples/imagenet/create_imagenet.sh

Creating train lmdb...

I0608 16:58:42.846923  8629 convert_imageset.cpp:86] Shuffling data

I0608 16:58:42.847262  8629 convert_imageset.cpp:89] A total of 0 images.

I0608 16:58:42.847405  8629 db_lmdb.cpp:35] Opened lmdb examples/image1000test200/ilsvrc12_train_lmdb

Creating val lmdb...

I0608 16:58:42.880640  8633 convert_imageset.cpp:86] Shuffling data

I0608 16:58:42.881011  8633 convert_imageset.cpp:89] A total of 0 images.

I0608 16:58:42.881208  8633 db_lmdb.cpp:35] Opened lmdb examples/image1000test200/ilsvrc12_val_lmdb

Done.

 

Can anyone offer assistance?

[Anne]

Hello,
You should double check the path to your images.

[Jalen Hawkins]

Thank you just like you said the path wasnt correct i was missing a '/' but i have now ran into a new issue.

I0610 14:14:56.620982  5936 layer_factory.hpp:77] Creating layer data

I0610 14:14:56.621083  5936 net.cpp:91] Creating Layer data

I0610 14:14:56.621088  5936 net.cpp:399] data -> data

I0610 14:14:56.621104  5936 net.cpp:399] data -> label

I0610 14:14:56.621111  5936 data_transformer.cpp:25] Loading mean file from: data/ilsvrc12/imagenet_mean1.binaryproto

I0610 14:14:56.621163  5940 db_lmdb.cpp:35] Opened lmdb examples/image1000test200/ilsvrc12_val_lmdb

terminate called after throwing an instance of 'std::bad_alloc'

  what():  std::bad_alloc

*** Aborted at 1465582496 (unix time) try "date -d @1465582496" if you are using GNU date ***

PC: @     0x7f99a1d89418 gsignal

*** SIGABRT (@0x3e800001730) received by PID 5936 (TID 0x7f9963b1a700) from PID 5936; stack trace: ***

    @     0x7f99a1d894a0 (unknown)

    @     0x7f99a1d89418 gsignal

    @     0x7f99a1d8b01a abort

    @     0x7f99a23c284d __gnu_cxx::__verbose_terminate_handler()

    @     0x7f99a23c06b6 (unknown)

    @     0x7f99a23c0701 std::terminate()

    @     0x7f99a23c0919 __cxa_throw

    @     0x7f99a23c0ebc operator new()

    @     0x7f99a2fc6cfd std::__cxx11::basic_string<>::_M_construct<>()

    @     0x7f99a2fc7c0b _ZN5caffe2db10LMDBCursor5valueB5cxx11Ev

    @     0x7f99a2fc21ce caffe::DataReader::Body::read_one()

    @     0x7f99a2fc267d caffe::DataReader::Body::InternalThreadEntry()

    @     0x7f99a01f75d5 (unknown)

    @     0x7f999faa56fa start_thread

    @     0x7f99a1e5ab5d clone

    @                0x0 (unknown)

Aborted (core dumped)

Can anyone help me with this?

[Anne]

Maybe this can help https://github.com/BVLC/caffe/issues/4124

[Jalen Hawkins]

 Im sorry i guess i skipped a few things. I have resolved the problem of having no files in the database. now i am trying to train the lmdb with ./build/tools/caffe train --solver=models/bvlc_reference_caffenet/solver1.prototxt

I adjusted the solver.prototxt file which is why the path says solver1.prototxt. but its giving me the error above. The training database appears to be working correctly but the vallidation database is relaying an error message.