lhmryan / deep-supervised-hashing-dsh Goto Github PK

Good day!

I cannot seem to generate the algorithm as is. The following error occurs at 64% (Building CXX object src/caffe/CMakeFiles/caffe.dir/layers/hashing_loss_layer.cpp.o)

[...]/include/caffe/util/device_alternate.hpp:15:36: error: no ‘void caffe::HashingLossLayer::Forward_gpu(const std::vector<caffe::Blob>&, const std::vector<caffe::Blob>&)’ member function declared in class ‘caffe::HashingLossLayer’
const vector<Blob>& top) { NO_GPU; }
^
[...]/src/caffe/layers/hashing_loss_layer.cpp:118:1: note: in expansion of macro ‘STUB_GPU’
STUB_GPU(HashingLossLayer);
^
[...]/include/caffe/util/device_alternate.hpp:19:39: error: no ‘void caffe::HashingLossLayer::Backward_gpu(const std::vector<caffe::Blob>&, const std::vector&, const std::vector<caffe::Blob>&)’ member function declared in class ‘caffe::HashingLossLayer’
const vector<Blob>& bottom) { NO_GPU; }
^
[...]/src/caffe/layers/hashing_loss_layer.cpp:118:1: note: in expansion of macro ‘STUB_GPU’
STUB_GPU(HashingLossLayer);
^
src/caffe/CMakeFiles/caffe.dir/build.make:2054: recipe for target 'src/caffe/CMakeFiles/caffe.dir/layers/hashing_loss_layer.cpp.o' failed
make[2]: *** [src/caffe/CMakeFiles/caffe.dir/layers/hashing_loss_layer.cpp.o] Error 1
make[2]: *** Waiting for unfinished jobs....
CMakeFiles/Makefile2:272: recipe for target 'src/caffe/CMakeFiles/caffe.dir/all' failed
make[1]: *** [src/caffe/CMakeFiles/caffe.dir/all] Error 2
Makefile:127: recipe for target 'all' failed
make: *** [all] Error 2

作者您好，
您在文章中计算loss的时候是和Siamese网络一样三个输入，您也解释了在线生成图片对的好处以及使用单层网可以达到和Siamese网相近的cost。
我想请问一下，您测试的时候是单张输入还是两张？

Dear authors,
when I am trying to run the code, I got the following error "error: ‘class caffe::LayerParameter’ has no member named ‘hashing_loss_param’". I wonder is this the problem of the caffe version? Would you please tell me which caffe you used?

I try to make this project on the VMware with Ubuntu without GPU and failed.
error info. is follow:
In file included from ./include/caffe/common.hpp:19:0,
from ./include/caffe/blob.hpp:8,
from ./include/caffe/layers/hashing_loss_layer.hpp:6,
from src/caffe/layers/hashing_loss_layer.cpp:4:
./include/caffe/util/device_alternate.hpp:15:36: error: no ‘void caffe::HashingLossLayer::Forward_gpu(const std::vector<caffe::Blob>&, const std::vector<caffe::Blob>&)’ member function declared in class ‘caffe::HashingLossLayer’
const vector<Blob>& top) { NO_GPU; }
^
src/caffe/layers/hashing_loss_layer.cpp:118:1: note: in expansion of macro ‘STUB_GPU’
STUB_GPU(HashingLossLayer);
^
./include/caffe/util/device_alternate.hpp:19:39: error: no ‘void caffe::HashingLossLayer::Backward_gpu(const std::vector<caffe::Blob>&, const std::vector&, const std::vector<caffe::Blob>&)’ member function declared in class ‘caffe::HashingLossLayer’
const vector<Blob>& bottom) { NO_GPU; }
^
src/caffe/layers/hashing_loss_layer.cpp:118:1: note: in expansion of macro ‘STUB_GPU’
STUB_GPU(HashingLossLayer);
^
Makefile:572: recipe for target '.build_release/src/caffe/layers/hashing_loss_layer.o' failed
make: *** [.build_release/src/caffe/layers/hashing_loss_layer.o] Error 1
make: *** 正在等待未完成的任务....

I think that without GPU cause this problem. Can I make this without GPU or should I change a device?
Looking forward to receive reply! THX!

Dear authors,
Could you please tell me how to get the NUS-WIDE data set ? I mean the data set that contains only 21 concepts. I've only found data sets with 81 concepts on the NUS-WIDE website. Thank you !
@lhmRyan

readme上说为了在线生成图像对，可是hashing_loss_layer不是已经实现了这个功能了吗？而且protext里也没有用这个data layer，您能解释下吗？

Dear authors,
when i follow the instructions on "http://caffe.berkeleyvision.org/installation.html" to compile the source code, when i set "USE_CUDNN := 1",i got the error: few arguements for cudnn::createPoolingDesc ,i compared the same file with caffe and copy the caffe's code and then compile this file successful. But when i continued compile other files also have this error and then i known these problems are caused by the version of CuDnn . So i set"#USE_CUDNN := 1",then i can "make pycaffe" and "make all" successful ,but when i want to "make test",i got the error "no matching function for call to ReadImageToDatum and ReadFileToDatum",i also compiled this file with caffe's and they are the same so i can't resolve this problem, can you give me some advise ?

dear:
I'm recreating your program of DSH, but I found that the hash loss layer does not have reverse-propagating code. Did you not published it ? Is it convenient for you to provide it? thank you a lot!

我看有的人用的是测试集图片查训练集图片。请问哪一个更标准一些？

你好，仔细阅读过你hashing_image_data_layer.cpp的代码，有个疑问，比如batch_size设200，cat_per_iters设为10，则每个类中取20个样本，之后如何构造相似/不相似样本对呢？因为你的代码中同一个类的20个样本，是顺序保存在cpu_data中的，之后读取的话是不是相似的样本对会明显多于非相似的？谢谢~

你好，你的代码在make test 时会报
make: *** [.build_release/src/caffe/test/test_memory_data_layer.o] Error 1
的错误，但是我直接编译官网的caffe不会出现这个问题，请教一下是为什么呢？

The default hash code dim is 12. I change it to 24 directly in train_test.prototxt instead of fine_tuning. Other hyperparameters remain unchanged. and I run train_full.sh and get a worse mAP of 0.6853 while mAP of 12 dim is 0.6987. I don‘’t know why . When I change it to 36. I get a mAP of 0.5839.

Could you give me some advise to improve the result when the hash dim is larger? Thx~

Hello,dear author,I am working on a project which needs hashcode to quickly retrieval. You have done a great job on deep hashing. I want to ask you a question,"How can I read the hashcode from the file 'code.dat'?"