how can train my own data ,I check the trainAll.m ,but it is alarmming all the time

[yatou1991]

I check the trainAll.m, but it is alarmming all the time.like this

what's the matter?can you help me?

[playerkk]

It means you do not have the file "train.txt", which specifies names of training images.

[yatou1991]

hallo,professor! May I ask you some question about some details of #the training progress. I am really curious about it.I read your paper about the DRFI method ,and read the code of it,but didn't understand the progress of the label building.

in your paper ,you said you connect each region and its spatially-neighboring regions forming a
set of pairsP = {(Ri,Rj)},and learn the probability p(ai = aj), where ai is the saliency label of the region
Ri. then Such a set of pairs into two parts: a positive part P+ = {(Ri,Rj)|ai = aj} and a negative part
P􀀀 = {(Ri,Rj)|ai ~= aj}.

the problem is that the saliency label be get ,I read the the relatted code and know you had segement the image to pieces and get the saliency code,and then change them to 1 or -1 to label ,but how can you change it like that ,this is the problem that I am being puzzled ,Could you help me to get it!! I'm really hope to get you help !And thank you to read my question

[playerkk]

I am not sure if I fully understand your question. The pairs you mentioned used to learn a binary classifier, indicating if two adjacent superpixels belonging to the same class. It is used ONLY for the training phase.

You can also read the journal version of the paper http://www.readcube.com/articles/10.1007/s11263-016-0977-3?author_access_token=wKZPweH_udFQ5vcd_NBsXPe4RwlQNchNByi7wbcMAY6bGK4nxWGm7p1_IwKxtD7cM55EhD3uI8Yr3Bf64FyCfVfipG0la-Rwo5Oh-ia9qwzPR7T5UM-fkSJaze6jVSoOQkq5dyl404AvLS9duq2kQw%3D%3D.

[yatou1991]

Thank you,professor Jiang!! I read the paper of DRFI method carefully,and understand the idea a little about the question I'm mentioned。And I will carefully read the paper that you advise me to read . I'm really thank for your help !!! But Can I ask one more question which about the Learning the regional saliency regressor ?

can you help me to understand the formula in the picture? the parameter of the Random Forest (f,t) can be get by this formula,but I read some book about the Random Forest Dicision Tree ,I found peopel usually get the best split by the information gain.If the formula represent the information gain? But I found the formula in the book is differet from the one in your paper ,so I'm a liitle puzzled,can you help me to know the reason，Thank you,professor Jiang!!!!

[yatou1991]

I am sorry ,I don't know If the picture which is in the above comment can show on your side ,so I send it again.

[playerkk]

It's a regression forest. The information gain is usually used for a classification forest.

[yatou1991]

Is there related conference? could you tell me? thank you,professor!