AI to Design Halloween Costumes

Artificial intelligence can be really spooky in the eve of Helloween
29 October 2018   789

Janelle Shane spoke about experimenting with AI on the eve of Halloween. The neural network was trained with a database with the names of costumes and it learned how to create its own. Janelle was able to do this with The New York Times: Editor Jessia Ma illustrated the results.

Last year, readers of the blog AIweirdness helped Janel to collect a base of 4.5 thousand names of suits. In 2018, she used the textgenrnn neural network and collected 7,100 samples using the New York Times publication (sent by people during the year). At the first stage, the algorithm created the word and compared it with examples from the database. In the event of a mismatch, the neural network changed the structure of the selection of letters. With each stage of the development of AI (they were called "epochs"), the generated costumes became more real to be realized.

In the first epoch, the costumes “Watand Hampir”, “Deadly Zanzai Vom” met. In the third, it was already “Greek beer” and “Darot Vader”. By the fifth stage, the AI ​​generated a “must-have minivan”, “Princess Laya”. At the seventh stage, a “giant box” and a “cyborg baby man” met. By the ninth, a “chewing cow” and a “wild Thor-pirate” appeared. And at the eleventh stage, the neural network created already full-fledged costumes, like the “death eater” or “witch hat”.

With the growing popularity of machine learning, the complexity of tasks grows and the scope of AI is expanding. At the end of October 2018, Honda, SoundHound, and three universities — Washington, Pennsylvania, and MIT — began to develop the Curious Minded Machine, an artificial, self-learning intelligence. Scientists expect the system to understand the actions of a person and offer him more effective ways to achieve goals.

Nvidia to Open SPADE Source Code

SPADE machine learning system creates realistic landscapes based on rough human sketches
15 April 2019   671

NVIDIA has released the source code for the SPADE machine learning system (GauGAN), which allows for the synthesis of realistic landscapes based on rough sketches, as well as training models associated with the project. The system was demonstrated in March at the GTC 2019 conference, but the code was published only yesterday. The developments are open under the non-free license CC BY-NC-SA 4.0 (Creative Commons Attribution-NonCommercial-ShareAlike 4.0), allowing use only for non-commercial purposes. The code is written in Python using the PyTorch framework.

Sketches are drawn up in the form of a segmented map that determines the placement of exemplary objects on the scene. The nature of the generated objects is set using color labels. For example, a blue fill turns into sky, blue into water, dark green into trees, light green into grass, light brown into stones, dark brown into mountains, gray into snow, a brown line into a road, and a blue line into the river. Additionally, based on the choice of reference images, the overall style of the composition and the time of day are determined. The proposed tool for creating virtual worlds can be useful to a wide range of specialists, from architects and urban planners to game developers and landscape designers.

Objects are synthesized by a generative-adversarial neural network (GAN), which, based on a schematic segmented map, creates realistic images by borrowing parts from a model previously trained on several million photographs. In contrast to the previously developed systems of image synthesis, the proposed method is based on the use of adaptive spatial transformation followed by transformation based on machine learning. Processing a segmented map instead of semantic markup allows you to achieve an exact match of the result and control the style.

To achieve realism, two competing neural networks are used: the generator and the discriminator (Discriminator). The generator generates images based on mixing elements of real photos, and the discriminator identifies possible deviations from real images. As a result, a feedback is formed, on the basis of which the generator begins to assemble more and more qualitative samples, until the discriminator ceases to distinguish them from the real ones.