machine learning - Convolutional Neural Networks with Caffe and NEGATIVE IMAGES -


when training set of classes (let's #clases (number of classes) = n) on caffe deep learning (or cnn framework) , make query caffemodel, % of probability of image ok.

so, let's take picture of similar class 1, , result:

1.- 90%

2.- 10%

rest... 0%

the problem is: when take random picture (for example of environment), i keep getting same result, 1 of class predominant (>90% probability) doesn't belong class.

so i'd hear opinions/answers people has experienced , have solved how deal no-sense inputs neural network.

my purposes are:

  1. train 1 more class negative images (like train_cascade).
  2. train 1 more class positive images in train set, , negative on val set.

but purposes don't have scientific base execute them, that's why ask question.

what do?

thank in advance.

rafael.


Comments

Post a Comment

Popular posts from this blog

toolbar - How to add link to user registration inside toobar in admin joomla 3 custom component -

im trying write component admin side , need add user creation. dont want duplicate existing codes add link. know how add task not simple link toolbar. thx help. if wanting button opens new user screen can try in component: $bar = jtoolbar::getinstance('toolbar'); $bar->appendbutton('link', 'users', 'new user','index.php?option=com_users&task=user.add'); you need add view.html.php file wherever you're creating toolbar. is need?
Read more

linux - disk space limitation when creating war file -

i have 1 gb disk space limitation hosting app , have used 700mb of it. however want create war file directory big (around 600mb) using command : jar -cvf root.war folderpath that's why encounter quota limitation , can't following. i'm looking way delete folder content when it's applying archive conserve disk space take note can't install zip tool in ssh shell ~ the code works on test directories, test first anyway, deletes files. it's not efficient. assumptions: 300mb free space, in destructively compress 600mb folderpath directory. none of files big -- if file in folderpath 300mb or larger, fail. notes: if ' $out ' file exists, nothing. first find pipes file list while , if ' $out ' doesn't exist in loop, create it, don't update -- if does exist, update it. each pass adds 1 more file, deletes unless file directory. after loop finishes delete empty directories too. unset u ; in=folderpath ; out=roo...
Read more