DMP’s AI recognition model for wire detection is an image recognition software that leverages deep learning technology. In general, when deep learning technology is applied to image recognition, high accuracy is achieved by preparing a large number of high-quality training images. On the other hand, there are many cases where it is not possible to prepare enough training images depending on the target of application. In the development of the AI recognition model for wire detection, since it is costly to collect a large amount of training image data, DMP used the Data Augmentation method which is to augment the amount of data from an existing data set using various techniques. By generating training data with different wire thicknesses, colors, and tints, DMP was able to develop a highly accurate AI recognition model without having to acquire and create additional data manually. DMP applies the Data Augmentation techniques, including Generative Adversarial Network (GAN), to the generation of training images. GAN consists of a generative model (Generator) that generates images and a discriminator model (Discriminator) that determines whether the image is real or created by the generative model. These two models compete in training to obtain a generated image that is closer to the real thing.

DMP strives to promote the spread of the AI recognition model for wire detection as a solution for customers who develop drones and autonomous mobile robots. DMP will also work on the further improvement of the functionality of this recognition model based on feedback from customers. Furthermore, DMP will develop and offer high-quality AI recognition models, not limited to the wire detection model, by leveraging the Data Augmentation techniques including GAN according to various customer applications and their requirements.