FV+CNN

[Former user]

Hello,

I try to execute the file run_experiment with dcnn (FV + CNN) on the dataset cub. the parameters are: setupNameList = {'dcnn'};
encoderList = {{dcnn}};

the accuracy generate is equal to 13.7 that is different that you mentioned in your paper. I do not understand why this difference.

Thank you for helping me to find the right accuracy.

[tsungyu]

I run the dcnn again using the code in this repository. I got the right accuracy instead of the number you mentioned. Let me know more detail about what you have done and what you have changed in the setting and then I might be able to identify the problem.

[amira fersi]

thanks for you.

i just modified the following lines : run experiments: setupNameList = {'rcnn', 'dcnn', 'bcnnmm'}; ----> setupNameList = {'dcnn'}; encoderList = {{rcnn}, {dcnn}, {bcnnmm}}; ---->encoderList = {{dcnn}};

model_train: opts.type = 'rcnn' ; ----> opts.type = 'dcnn' ;

model_setup: opts.encoders = {struct('type', 'rcnn', 'opts', {})} ;---->opts.encoders = {struct('type', 'dcnn', 'opts', {})} ;

[tsungyu]

You can only edit the run_experiments.m and those change will overwrite the default setting in model_train and model_setup.

Other than this, I am wondering if you pulled the latest version? The line opts.encoders = {struct('type', 'rcnn', 'opts', {})} has been commented out after the commit 0318682.

[amira fersi]

Thanks ,it is ok