How to build your own dataset with Neural Network Console? (Part 4)
Deep learning has become more and more vitally important in our daily life. The way that we unlock our phone and the intelligent LEDs on the streets are both demonstrating the example of implementing the AI in terms of image recognition. There are many open sources on the internet that are about AI, including handwritten digit recognition, but how could we build our own dataset? In this article, I am sharing an application with you all, the Neural Network Console.
This article will be divided into 4 parts, first preparation of the datasets, second build the Network model, third create the dataset and last how to evaluate the model.
In part 3 we have delivered an okay model; but how could we improve that model?
The previous Learning curve and the Confusion matrix showed that label 1 is less than 90%, so how do we improve the performance?
We could improve the performance in 3 Steps:
Firstly, increase the data size and remove the duplicated data.
Secondly, increase the data size of the confused label, for example, label 3 got 22 label 1 and label 5 got 83, so we could increase the data size for both of them
Lastly, put all the label 1 removed data into the label that gave nothing.
Repeat part 1-3 and see what happens.
So, let's see how this performed.
First, label 3 got 20, label 1 less, and label 5 got 52 less.
Second, label 1 is back to 90%.
Apart from image recognition, the neural network console also supports Image generation and Image Segmentation.
Could the neural network console recognise faces? Yes! It would be interesting using this app as a starter.
Inputting different data sets and the various results generated.