1

I wanted to develop an app similar to this on my own which is based on image classification using neural nets. While in early phases of my research I came across through terms Deep Neural Nets and Convolutional Neural Nets.

What is the difference between two and which will be the best bet for me to work on for this project?

Edit1:- I looked on kaggle's digit recogniser for my initial research

radbrawler
  • 111
  • 1
  • 1
  • 4
  • [DNN](https://www.kaggle.com/kakauandme/digit-recognizer/tensorflow-deep-nn) and [CNN](https://www.kaggle.com/kobakhit/digit-recognizer/digit-recognizer-in-python-using-cnn) – radbrawler Jan 16 '17 at 16:14
  • I can't add more than two links. I'm short of reputation. – radbrawler Jan 16 '17 at 16:15

1 Answers1

0

The term deep neural nets refers to any neural network with several hidden layers. Convolutional neural nets are a specific type of deep neural net which are especially useful for image recognition. Specifically, convolutional neural nets use convolutional and pooling layers, which reflect the translation-invariant nature of most images. For your problem, CNNs would work better than generic DNNs since they implicitly capture the structure of images.

If you want a good layman's introduction to deep learning and CNNs, there's a great Nature paper (if you search "deep learning nature" on Google, it's one of the top links).

liangjy
  • 646
  • 4
  • 12
  • Does that mean I can use both .? – radbrawler Jan 16 '17 at 16:53
  • You can use either a generic DNN or a CNN, but a CNN will almost certainly give you better results. You should start out with implementing a DNN though, since it's easier and you'll gain some knowledge and intuition about neural networks. – liangjy Jan 16 '17 at 17:03