lelan-li / ssah Goto Github PK

I am intested in your contribution, but when run your repo, there is something wrong with the FLICKR-25K.mat:
KeyError: Unable to open object (object 'images' doesn't exist)

I found that the FLICKR-25K.mat are generated automatically.

But when i download the file which your data.md mentioned, and put it under the ./data,I found it have another problem:

OSError: Unable to create file(unable to open file: name="./data/FLICKR-25K.mat",errno=17,error message='File exists', flages=15,o_flags=c2),

so could U please tell me how to run your code?

Hi,
I ran your code and it can re-produce the performance on flickr, but fails on nus-wide.
Did you use different parameters ? can you provide them?
thanks

hi,
I ran your code, a epoch takes several hours.
my gpu is RTX2070.
thanks!

in Mirflickr25K, the map of 64 bit hash is 80 80, but i get this

I have plotted two Precision-Recall curves on the result of Flickr-25K. Some settings as followed: bit=16, using vgg19 features. The codes and result are provided in SHARING.

These curves look a bit strange because some of their starting points don't show a downward trend as those curves shown in DCMH and SSAH.

Is there a problem with my code?

halo i have an indecerror and i take a screenshot. what should i do to resolve the problem?Thank you !

I re-run the experiment based on VGG-19 backbone network. However, the results on VGG-19 is rather far less than the ones reported on the original paper. It might be caused by the hyperparameters setting?
If so, then how to set the parameters different from those based on CNN-F.
Thank you very much!

could U please tell me which version of python and tensorflow your code runs?

Is it the real code used to implement your experiments?
I have found the loss in your code is not consistent with the description in your paper and there are many matrix dimension errors in your batch training phase.
e.g.

hi, your work is great. Although there are some differences with the original paper.
can I get your pretrained vgg16.mat?thanks
Besides, your tensorflow version is ?