This is mi final project of Ironhack data analysis bootcamp In this project I have created a script with a graphical interface where a user can create a work of art through his feelings.
It has been a challenge to program all this with python. Creating art with artificial intelligence brings art closer to the people. Using different technologies learned in Ironhack, it has been like doing many projects in one.
First, the interface asks the user to say what he or she wants to express in order to generate the work. The interface collects the voice, records it and analyses it with machine learning algorithms. It also analyses the context of the user's sentence using neural networks (NLP). On the one hand, it generates a style image. This style image depends on the context of what the user says, the type of message it is (positive, negative) and the emotion (classified by the tone of voice) with which it is said (happy, sad, angry...). Based on these variables analysed with machine learning, nlp and Word2Vec (two-layer neural network), the programme chooses a tone, rearranges the vector generated by the neural network and generates an np.array that it converts into an image.
Then, the programme asks the user to press the second button to start the second phase. In the second phase, forms will be associated with how the user has said the message (chosen according to the sentiment classification performed by the ML model).
Finally, we use an already trained deep convolutional neural network VGG19 to which, after processing, we pass the following input:
- An initial image (the chosen shapes)
- Intermediate layers of this image
- Intermediate layers of the image obtained from the np.array generated by the context.
With all this, the neural network returns to the output a style transfer that looks as follows.
- Beautiful Soup
- Word2vec (natural language processing)
- Sklearn (machine learning algorithms)
- Tensorflow
- Keras - Vgg19 (Deep convolutional neural networks)
- Tkinter (Python interface)