Hacker News Reader

Comment by 3rd3 | Hacker News Reader

Are there actually any image feature detectors and descriptors involved (like blob, edge and texture detectors) or is this solely based on artificial neural networks?

View on HN

Interestingly it has been shown that the result from some neural networks is equivalent to using classification with some predefined filters. These filters could be considered as a feature descriptor. See this talk from CVPR http://techtalks.tv/talks/plenary-talk-are-deep-networks-a-s....

Thanks for sharing this. I enjoy Mallat's point of view. He has some similar talks on videolectures.net for anyone who's interested.

AFAIK, it's using a Deep Neural Network; which means, the inputs are, basically, pixel values (possibly normalized), and all feature detection, etc. is done in the layers of the network.

yep, they try to learn an image's high level features by learning an autoencoder (that is a transform that takes an image and tries to produce the same image) via a sandglass shape multi layer network. Here is a very readable paper by Hinton himself that describes the approach:

http://www.cs.toronto.edu/~hinton/science.pdf

Could it maybe be worthwhile to augment the data with simple image features? E.g. the human visual system is believed to rely on high-level/top down as well as on local/bottom up features (although that might also be simply because of the necessity to compress things for the low nerve count in the optical nerve).

A Deep Net (to be specific: a deep belief network which is a series of stacked RBMs, not Stacked Denoising AutoEncoders for clarification that there's a difference) usually can benefit from a moving window approach (slicing up an image in to chunks) to simulate a convolutional net. This can help a deep net generalize better.

That being said: even deep learning requires some sort of feature engineering at times (even if its pretty good with either hessian free training or pretraining).

The main thing with images is ensuring scaling them.

The trick with deep belief networks in particular is to make sure the RBMs have the right visible and hidden units (Hinton recommends Gaussian Visible, Rectified Linear Hidden).

Happy to answer other questions as well!

I think it is a convolutional network trained only with gradient descent, since pressing source code links to convnet project.

I'm pretty sure there's not an autoencoder involved, it just looks like a vanilla conv net.

This is the implementation: http://torontodeeplearning.github.io/convnet/

Reply on news.ycombinator.com