Artificial Intelligence - From Childish Scribbling Art to Realistic Natural Landscape - Deep Learning of GAN

Deep Learning, formally a subset of Machine Learning used in big data structures and also in neural networks.

The range of development since the introduction of neural networks which is based on the concept of mimicking human neuron is huge and really opened many new scope of solving the problems.

Here comes one such surprise,

NVIDIA research group built a system, which transforms a doodle painting of a childish scribbling into a photo realistic landscapes.

This system is built on GauGAN - GAN's deep learning algorithm (generative adversarial networks).

GAN is one of the best deep learning algorithm considered as featured idea in the last 10 years - as it solved the biggest huddle of deep learning system with reducing the training time and also extending the possiblities more than the available data.

GAN consists of a Generator and Discriminator, Generator proves the possible output and discriminator creates the new possibility based on the generated result extends the feature of data.

For example, during the training of data( taking cat images at this instance), generator will predict and discriminator will feed back another fake cat data based on the generator.Thus, will make training of data will be fast forwarded and also test data will be more accurate respect to the time.

Coming back to the GauGAN network that has been trained on a million real world images on natural resources producing output on similar scale.

The model is well equipped as it finds water will reflect the other objects in the frame and also sun rise, sun set on giving the time frame of effect.

Find more here,

Post a Comment