Please scan QR code by WeChat and
completed the registration
Dear Valued Guest:
We cordially invite you to attend the Intel® AI DevCon (Intel® AIDC) from November 14-15, 2018 at the China World Summit Wing in Beijing. This world-class AI conference will bring together over 2,000 of the world's most influential AI experts, industry leaders, data architects, and technical engineers. Over 30 industry partners will showcase cutting-edge solutions, products, and technological applications in the AI sector. This conference will focus on technological frontiers, trends, opportunities and challenges related to AI. This will be a wonderful opportunity for you to meet face-to-face with experts in the field and help create the world's top platform for AI partnerships.
At the Intel® AIDC 2018 we will explore cutting edge topics in AI with leading experts in the field, while offering a unique conference experience. We hereby invite you to attend this summit, and hope you will join us in our journey to drive the transformation of industry through AI technology.
Cordially Invited by Intel China
September 2018
Intel® Myriad™ X third generation VPU delivers world class-leading performance in computer vision and deep neural network inferencing.
AI accelerator demo: one or multi-ways video streaming input via USB is inferenced with typical neural network algorithms. Human bodies are detected and identified, the frame rate is about 60 fps.
High performance Smart Camera demo: Camera input to Myriad™ X’s own ISP; through image pre-processing, objects detected by mobileNetSS, object recognized by mobileNet and output NN analysis result to display on PC via USB, the frame rate is about 30 fps.
Intel ® Movidius ™ Myriad ™ X based Vision Accelerator is a product that provides powerful deep neural network inference capabilities for fast and accurate video processing together with Intel OpenVINO ™ toolkit, to meet the needs of edge computer vision applications and help solution providers and their customers take full advantage of video analysis in more practical scenarios .
Ferrari Challenge is a single model race series. Intel® is applying AI to the drone-based race coverage. Solution uses fine grain classification and few shot learning to match each car to its driver.
Coach is an open source Python* RL framework for AI agent development and training. It supports many state-of-the-art RL algorithms, integrated with multiple simulation environments.
OpenVINO™ is an Intel® FREE toolkit to fast-track high-performance computer vision and deep learning inference to vision applications. In this demo we show OpenVINO™ deep learning application running on Intel® FPGA PAC board (Arria® 10), which uses deep learning for Vehicle and Chinese License Plates Detector, Vehicle Color/Type Recognition and Chinese License Plates Recognition.
Introduce the why, what and how of the Unified Big Data Analytics and AI Platform - Analytics Zoo, which is based on Apache Spark and BigDL™. This demo shows example source code of Analytics Zoo to achieve practical demo such as “Soccer Star Recognition”, “Image generation”, etc.
DL Boost is a set of processor technologies designed to accelerate AI deep learning use cases. It extends Intel® AVX-512 with a new Vector Neural Network Instruction.11X speed up in performance delivers a better TCO for customers and a more consistent experience for end users.
Introducing a New Generation of Versatile, Productive Plug-and-Play AI development kit, powered by the Intel® Movidius™ Myriad™ X VPU.Delivering up to 8X the performance of its predecessor.
PERFORMANCE IMPROVEMENT OF TRAINING TIME OF M-CNN USING HIGH-CONTENT IMAGING.
Supported by Inspur's NF5280M5 AI Server with 4 F10A FPGA cards, this live demo shows FPGA-based acceleration for real-time image classification is enabling high efficiency, low latency, and high performance image recognition with high accuracy. Inspur F10A is equipped with Intel® Arria™ 10 FPGA.
Corerain AI Edge Solution uses Intel Arria™ 10 FPGA 660 SoC and Corerain self-developed chip structure, CAISA, providing powerful AI real time detection solution, which can support 16 full HD surveillance cameras to detect faces, cars, and traffic situation with the power consumption of 36W.
Horizon Robotics 360° visual perception system consists of three sets of Matrix that are in-house developed by ourselves. By designing the position of mounted cameras, our computing platform for autonomous driving can achieve a 360-degree perception around the vehicle.
XForce is our newly launched FPGA Edge AI computing platform based on Horizon's BPU 2.0 AI Processor architecture. With the hardware deeply customized to seamlessly run our proprietary algorithms, XForce provides superior visual perception capabilities with extraordinary efficiency in cost and power consumption compared with generic GPU solutions. XForce now supports functionalities of human behavior analysis and video structuring.
Run deep learning jobs like face recognition, object detection and word vector training using Intel Xeon ®Scalable Processor and Intel Optimized Caffe framework on QingCloud deep learning platform.
SIGHTA is a real-time traffic AI analysis product based on video algorithms and computer vision techs. The core features includes traffic data collection, traffic accidents detection, device status evaluation and highway fog detection.
Through UAI platform, we can help customers to achieve AI training tasks and inference services.
JD Magic Cam, a smart camera developed by China's largest retailer, JD.com that allows people to superimpose their images on different backdrops in real-time, provides a unique marketing solution for offline stores. A real-time mapping function enables users to virtually switch instantly from one scene to the next. Users can add 2D and 3D special effects to each scene, and can seamlessly share images created with JD Magic Cam across social media with friends to aid in purchasing decisions. In addition to providing a source of entertainment for consumers, JD Magic Cam also empowers businesses with enhanced marketing and user engagement capabilities. JD Magic Cam is powered by Intel® i5-6500 processor. expand and upgrade future innovations.
The whole solution is based on huge volume of image data captured on the factory production line in real time which are later pre-processed & sent cloud (Private or public). In the cloud, the AI engines are trained with the data on Xeon Scalable processors. Finally, the inference engine can be implemented on either the cloud side or the front-end Core® i7, depending on the network condition and the latency requirement, to conduct various kinds of vision inspections.
Skysys is the very first team ever in China that is engaged in the development of drones' hangar, aka the Drone Autonomous Operating Platform or the Drone-enabling System. The advent of this particular system, which is named as UltraHive, has offered a one-stop solution that resolves all the drawbacks of current drones, including difficult control, short endurance, lack of hangar and poor universalization.
Sketch Guess This demo presents a Google mini-game 'Quick, Draw!' clone made using HS and Python There are 350 stages in total. In each stage you will be asked to draw a specific object on the paper The neural network in HS will decide if your drawing matches the object's title.
Face Recognition This demo presents a realtime multi-face detection and recognition system that is built with HS, It runs on most platforms with nearly no extra computational cost on the host The system can also register new faces on the fly
Tarsier Module by VIONVISION, based on the unparalleled small-sized Movidius™ chip with low power consumption (about 1.5W) but powerful computing power. Smart Cameras based on Tarsier Module:Smart Face People Counting Camera, All-in-one People Counting Camera,Smart Fisheye Camera.
Based on the Intel® Xeon® SP Platform, Dr. Pecker greatly improves the efficiency and accuracy of medical imaging diagnosis, and reduces the rate of missed diagnosis and misdiagnosis.
Intel and HuiyiHuiying jointly developed the‘Artificial Intelligence Breast Cancer Full Cycle Health Management Platform’, which covers the entire process from breast cancer prevention screening to follow-up monitoring, providing a reliable end-to-end artificial intelligence solution to hospitals.
Currently,with introduction of 4G and coming 5G,huge number of image and videos contents are generated continuously in the internet, one major issue of the content-provider is to realize an effective detection to avoid unpermitted contents publicated , We demonstrate an effective AI based image detection using Intel® AIO PAC card for certain content. Combining with Vismarty's own AI algorithm and FPGA implementation, an outperformed processing speed will be demonstrated comparing with other solutions.
4Paradigm Prophet AutoML is a low-threshold, automated AI application building tool.4Paradigm AIO's software and hardware integrated design is a genuine combination of the advantages of 4Paradigm's state-of-the-art AutoML technology and Intel® Xeon® SP's advanced hardware architecture.
The edge-to-cloud Industrial Edge Computing platform, train the AI model in cloud and run influence at edge.
CTC-Intel® Joint Lab demonstrate AI related achievement included AIaaS platform and industry-vertical solutions like 'Intelligent police, 'Intelligent farm' and etc.
Intel® provides end-to-end AI solution for wild tiger protection Intelligent camera powered by Intel Movidius is used for onsite recognition of wild animals On cloud, tiger re-identification based on Intel® Xeon® is used for individual tiger’s data extraction and tracing.
Intel® Xeon® help reconstruct high resolution 3D model for Jiankou Great Wall based on the air survey by Intel Falcon 8+ drone Cracks and collapses on the Wall are automatically detected 3D-GAN network generate the missing parts of the Wall This is the first time that AI has been applied in large-scale ancient construction (digital) repair.
Intel® provides end-to-end AI solution for wild tiger protection Intelligent camera powered by Intel Movidius is used for onsite recognition of wild animals On cloud, tiger re-identification based on Intel® Xeon® is used for individual tiger's data extraction and tracing.
The above information will be continuously updated.
The final information will be finalised during the actual event day.
PRESENTER | POSTER TITLE IN ENGLISH |
---|---|
Anbang Yao (ILC) | Efficient Semantic Scene Completion Network with Spatial Group Convolution |
Yiwen Guo (ILC) | Deep Defense: Training DNNs with Improved Adversarial Robustness (to appear in NIPS'18) |
Xiaofan Xu (Movidius) | Hybrid Pruning: Thinner Sparse Networks for Fast Inference on Edge Devices |
Haim Barad (Intel AIPG Inference Group) | Early Exit for Fast Inference Classification |
Karthik Vadla, G Anthony Reina (AI Lab) | HOROVOD DISTRIBUTED TRAINING ON K8S USING MLT |
Yinpeng Dong (THU) | Boosting Adversarial Attacks with Momentum |
Chenyang Si (CASIA) | Skeleton-Based Action Recognition with Spatial Reasoning and Temporal Stack Learning |
Bo Peng (CASIA) | Position Determines Perspective: Investigating Perspective Distortion for Image Forensics of Faces |
Jiangyan Yi (CASIA) | Convolutional Interaction Network for Natural Language Inference |
Jingjing Gong (Fudan) | Convolutional Interaction Network fo Natural Language Inference |
Youmeng Li (TJU) | A Hydrology Simulation Method Using LSTM |
Shuai Gao (USTC) | SATB-Nets:Training Deep Neural Networks with Segmented Asymmetric Ternary and Binary Weights |
Jun Shi (USTC) | DSU-Net: Cervical Cancer CTV Automatic Labelling Method Based on Convolutional Neural Network |
Zhicheng He (NankaiU) | Content to Node: Self-Translation Network Embedding |
Wei Feng (TJU) | Active Camera Relocalization from a Single Reference Image without Hand-Eye Calibration |
Xin Yang (DLUT) | Active Matting |
Yihua Huang (NanjingU) | Reinforcement Learning-Based Automatic Machine Learning (AutoML) Algorithm & Framework |
Hao Wang (RenminU) | Laser Scar Detection in Fundus Images using Convolutional Neural Networks |
Sanyuan Zhao | Pyramid Dilated Deeper ConvLSTM for Video Salient Object Detection |
Li Jiang (SHJ) | ReCom: An Efficient Resistive Accelerator for Compressed Deep Neural Networks |
The above information will be continuously updated.
The final information will be finalised during the actual event day.
Please scan QR code by WeChat and
completed the registration