For a research project, I need to do a segmentation on images. Since the motivation is nothing any of the big networks was ever trained on, I would ask if it still makes sense to use pretrained segmentation networks like SegNet to do the task. (Maybe by giving the network just a few (1 - 50) training examples?)
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
For research project point of view, I would prefer to use own training examples instead of pre trained images as it would be beneficial how you manipulate and segment the images according to the project's bottom line. Then convert it to Convolution Net type. Note: while using your own examples keep the image in fixed frame size as it would be very helpful for image processing.
Hope this helps.
-
$\begingroup$ How many trainings example would i need though? And should i still use a pretrained Network (since it already has learned all the basic features in the bottom layers). $\endgroup$ Jun 12, 2017 at 6:41