Intel to Present Neural Compute Stick 2

Neural Compute Stick 2 is an autonomous neural network on a USB drive
15 November 2018   1140

At the Beijing conference, Intel introduced Neural Compute Stick 2, a device that facilitates the development of smart software for peripheral devices. These include not only network equipment, but also IoT systems, video cameras, industrial robots, medical systems and drones. The solution is intended primarily for projects that use computer vision.

Neural Compute Stick 2 is an autonomous neural network on a USB drive and should speed up and simplify the development of software for peripheral devices by transferring most of the computation needed for learning to the specialized Intel Movidius Myriad X processor. Neural Compute Engine, responsible for the high-speed neural network of deep learning.

The first Neural Compute Stick was created by Movidius, which was acquired by Intel in 2016. The second version is 8 times faster than the first one and can work on Linux OS. The device is connected via a USB interface to a PC, laptop or peripheral device.

Intel said that Intel NCS 2 allows to quickly create, configure and test prototypes of neural networks with deep learning. Calculations in the cloud and even access to the Internet for this is not needed.

The module with a neural network has already been released for sale at a price of $ 99. Even before the start of sales, some developers got access to Intel NCS 2. With its help, projects such as Clean Water AI, which use machine vision with a microscope to detect harmful bacteria in water, BlueScan AI, scanning the skin for signs of melanoma, and ASL Classification, real-time translates sign language into text.

Over the Movidius Myriad X VPU, Intel worked with Microsoft, which was announced at the Developer Day conference in March 2018. The AI ​​platform is expected to appear in upcoming Windows updates.

Nvidia to Open SPADE Source Code

SPADE machine learning system creates realistic landscapes based on rough human sketches
15 April 2019   674

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.