 |
| The Alibaba Cloud booth at the NVIDIA AI Conference. |
NVIDIA's first artificial intelligence (AI)-focused regional conference in Singapore in late October featured state-of-the-art AI demonstrations and showcased startups in the company's Inception programme. Highlights of the
NVIDIA AI Conference showed how computing powered by graphics processing units (GPUs) could change the way tasks are accomplished:
 |
| The NVIDIA Tesla V100 is a data centre GPU and designed to accelerate AI, high performance computing (HPC) and graphics. Likened to offering the performance of 100 CPUs in a single GPU, it is powered by the NVIDIA Volta GPU architecture. Volta uses 640 Tensor cores to deliver a 5x increase compared to the prior generation NVIDIA Pascal architecture. Volta also relies on next-generation NVIDIA NVLink high-speed interconnect technology to delivers 2x the throughput compared to the previous generation of NVLink. |
 |
| The NVIDIA DGX Station is billed as a personal
supercomputer for AI development, and can also be used for workgroups. It is built on the same NVIDIA
GPU Cloud Deep Learning Stack powering all NVIDIA DGX Systems, and runs on four NVIDIA Tesla V100 GPUs, the NVIDIA NVLink, and new Tensor Core architecture. |
 |
| In the "AI imitating art" demonstration, real-time pictures captured by the camera at the top right are translated into artworks in the style of an example, in this case a painting by Van Gogh. The system does not apply a filter in the same way Photoshop might work - it knows where the people's faces are, and treats them differently. Note that this 'seed' painting produced a different artwork later on. |
 |
| In the "AI imitating art" demonstration, real-time pictures captured by
the camera at the top right are translated into artworks in the style of
an example, in this case a painting by Van Gogh. Note that this 'seed'
painting produced a different artwork earlier on. |
 |
| In the "AI imitating art" demonstration, real-time pictures captured by
the camera at the top right are translated into artworks in the style of
an example by Paul Cézanne at the bottom right. |
|
|
 |
| In the "AI imitating art" demonstration, real-time pictures captured by
the camera at the top right are translated into artworks in the style of
an example or 'seed' at the bottom right. |
 |
NVIDIA Metropolis provides the tools, technologies, and expertise to create and run smart cities. AI is used to capture, inspect, and analyse data to influence municipal services such as public
safety, traffic, parking management and law enforcement. NVIDIA DeepStream can simplify the development of scalable intelligent video analytics applications, such as determining if people in moving vehicles are wearing seatbelts, or where there are empty carpark spaces.
"If theres no safety there's no city to build a smart on top," noted Deepu Talla, VP and GM, intelligent machines at NVIDIA. He explained that 75% of the data traffic in cities is video data. There is so much of it that humans simply cannot process it all, and only review it after something has happened. "Traditionally video analytics has not been accurate, with many false positives (editor's note: saying something has happened when it has not)," he said. "Because of AI the accuracy is as good or better than humans." |
 |
| This demonstration shows how the NVIDIA Drive PX platform analyses road conditions, determining which part of the scene can be driven on, and where obstacles might be. |
 |
| This technology is most recognisable as the "animoji" capability in the iPhone X, but animates the entire face for more realistic telepresence experiences. This face can repeat a phrase recorded by visitors. |
 |
| This demonstration shows how the Jetson TX2, a very low power system can be used to identify the colour and type of car from an image or video of cars in a car park, even if the cars are at different angles and partially obscured. It can also identify licence plates in real time. |
 |
| This demonstration showed generative adversarial networks (GANS) working to create photorealistic faces. The GANS framework is the latest AI technique to sweep the industry. The method has two neural networks compete with each other. One tries to accomplish a task and the other tries to detect when the task is not done properly. This method has proven to help neural networks improve more quickly. |
The company is also working with 1,900 startups through the Inception programme, which provides startups with the technologies and other resources they need. Some of the AI startups in the Inception programme were on the showfloor, including:
 |
| RobotLinker changes develops complex robotic software using technology from the industrial Internet of Things, artifical intelligence, edge and cloud computing. One of its demonstrations included superimposing stick figures onto people in videos in real-time to show how their bodies move. |
 |
| RobotLinker's software tracks body silhouettes and can create articulated "stick figures" showing how bodies move. |
Abeja makes use of data collected by Internet of Things sensors to develop insights that help businesses. For instance, Abeja has used customer behaviour and point of sale data to forecast how long a transaction takes at the cashier, allowing the staff to adjust the amount of time they spent at the checkout area.
Adaline reads medical images and identifies breast cancer more accurately and consistently than humans. The company is working with hospitals to increase the accuracy of the system, which was requested by radiologists.
 |
| Asia Genomics uses deep learning to calculate biological age based on biomarkers in the blood. |
 |
| Dathena uses
machine learning and AI technologies to categorise documents in an
organisation, then helps to protect, retain or delete them according to
customisable business-centric rules. |
 |
| Graymatics uses HPE hardware to conduct cognitive multimedia analytics, basically transforming CCTV video into usable data. The company already has customers around the world. For a retail setting, Graymatics technology can identify the number of people visiting a store, broken down by time spent, gender, age, even the route taken through the store. Useful insights that can optimise sales might include reports on whether whether the target group for a product goes near the product, for instance. For surveillance, the technology can identify security anomalies or illegal parking and send alerts to the authorities. |
 |
| Hanalytics uses big data analytics and AI to uncover insights and to create predictive applications in the healthcare sector. |
Nodeflux can power smart city and retail applications with both traditional and cognitive computing.
Shentilium Technologies offers a tool to help brokerages make better hedging decisions as well as analytics tools for applications like cohort analysis - understanding customer profiles and identifying people with similar profiles who might also be interested in buying the same products.
Twistcode built a GPU-based supercomputer to accelerate a range of applications, including genomic data processing, rendering, deep learning and AI, finance, scientific and engineering simulations.
 |
| ViSenze
delivers intelligent image recognition solutions that convert images
into product search opportunities. Its technology is being used by
retailers like Rakuten, Uniqlo, ASOS and Flipkart. |
 |
| Identifying sushi types based on images. Xjera Labs focuses on developing AI-based video and image content analytics solutions. Key value propositions include multichannel analytics, minimal calibration required, and 20%-30% lower hardware specifications compared to other video analytics vendors. |
 |
Other exhibitors on the showfloor included Lenovo, which showcased a proof-of-concept 3D medical diagnostic system currently piloting at two hospitals in Beijing, China. The system is designed to assist the radiologist by screening scans and providing a preliminary diagnosis of the problem.
|
 |
| Imagene Labs is a Singapore-headquartered genetics-driven wellness company. Its Krease app uses a selfie to analyse facial wrinkles and simulate how facial skin will respond to various treatments. |
 |
| Kinetica uses the GPU to offer both business intelligence (BI) and AI on the same database platform. Use cases include industries which use predictive analytics, for example for fraud detection or pattern regonition. Testing for manufacturers can be completed in real time as can risk modelling for banks. |
Explore:
The
NVIDIA Inception Program is a global virtual accelerator programme that
offers access to technology, experts and training. Members receive
access to NVIDIA's AI computing technology and can reach out to a
network of deep learning experts and thought leaders; global marketing
support; introductions to Fortune 1000 companies and funding
opportunities, training from the NVIDIA Deep Learning Institute. Apply
Explore:
Read the
TechTrade Asia blog posts on:
NVIDIA's AI strategy
How far AI can go: the Alibaba experience
Singapore's approach to AI
The EDB-NVIDIA Future Talents Program
The shared NVIDIA and NSCC platform for AI initiatives in Singapore
How XjeraLabs uses NVIDIA Tesla technology to offer unique analytics solutions
ViSenze on MasterCard's Start Path
Hashtag: #NVAICONFERENCE17
No comments:
Post a Comment