Stick figure recognition and generation method based on Raspberry Pi and recurrent neural network

Abstract

The invention discloses a stick figure recognition and generation method based on Raspberry Pi and a recurrent neural network. The method comprises the following steps: (1) collecting a vector stick figure data set; (2) establishing a mathematical model of vector stick figure data; (3) preprocessing vector stick figure data; (4) constructing a recurrent neural network vector stick figure generation model based on a sequence-to-sequence model and a long-term and short-term memory network; (5) training the recurrent neural network vector stick figure generation model; (6) deploying the trained model on the Raspberry Pi; (7) collecting user voice by using a microphone, understanding user semantics by using a Google voice-to-text module and a natural language processing module, mapping the user semantics into a stick figure generation model as input, and controlling a printer to output a generated stick figure picture through Raspberry Pi. According to the method, the problem that vector data are difficult to process in an existing image generation method is solved, and the limitation of an end-to-end system for realizing language-to-image is solved.

Description

technical field [0001] The invention belongs to the technical field of image generation, and in particular relates to a stick figure recognition and generation method based on a raspberry pie and a recurrent neural network. Background technique [0002] The study of generative models is a very important and active research topic in the field of deep learning. Many effective generative models have been proposed in the current field of deep learning, such as NADE, variational autoencoder, DRAW, etc. In the field of image generation, the most popular method is generative adversarial network. The GAN model trains a generator and a recognizer at the same time, the generator is used to capture the data distribution to generate new samples, and the recognizer is used to distinguish real samples from generated samples. To better simulate the generative process, GAN trains both the generator and the recognizer in a zero-sum game framework. Through the zero-sum game and the conditio...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above-mentioned deficiencies in the prior art, the present invention provides a kind of stick figure recognition and generation method based on raspberry pie and recurrent

Images