Training a Yes/No NN for Image Classes



I am fairly a newbie to Neural Networks.

I wanted to ask if it is possible to train a NN to identify only one type object? For instance, a table from a large set of images, where the NN should be able to identify if new images are tables.

If yes can you please guide me in the direction to get started?


I wanted to ask if it is possible to train a NN to identify only one type object?...

so my understanding as of now is that if i train an NN on one class and one class alone with a fairly large amount of data then i can get it trained in a small amount of time rather than training it on a huge amount of data and consuming a larger amount of data...

Objective is to train a large no of such smaller NNs to create an ensemble of NNs.


Posted 2018-02-16T10:17:09.810

Reputation: 375

3Of course. Just label images with tables in them as 'yes' and images without tables in them as 'no' – Daniel – 2018-02-17T23:55:55.760



Yes, you can train a NN to detect only one type of object like a table. However, you probably will not want to train such a NN from scratch by showing some examples of tables and non-tables. You will need to use transfer learning on a model already trained on several image classes and teach it to also recognize your new class. This transfer learning requires a smaller set of desired images. You may need to give it some negative examples also. You should explore transfer learning with mobilenet, inception, and other pre-trained Tensorflow models if you are willing to use Python and Tensorflow

Adnan S

Posted 2018-02-16T10:17:09.810

Reputation: 253