News for hackers. Stuff that matters.
Re:How to trick a neural network into thinking a panda is a ... by dgacmu on Thursday December 24, 2015 @08:36AM (#10788096)
The Imagenet-based networks already have bounding boxes to identify the region of interest. It's important, because many of the training images have multiple "things" in them, but in the '12 dataset, only one label. It's not as much "feature engineering" as "object discrimination in the training set".
Re:How to trick a neural network into thinking a panda is a ... by argonaut on Thursday December 24, 2015 @09:05AM (#10788165)
That's not feature engineering, that's just a label for the task of interest (object localization).

Blanking out areas that are "not of interest" I would consider substantial feature engineering (unless the task you were training a net for was explicitly to find interesting vs. uninteresting areas).