DAMO Academy's official website is officially launched, revealing the architecture of Alibaba's city brain

In October last year, Alibaba announced the establishment of the "Damo Academy", implementing the dean responsibility system, with Alibaba Group CTO Zhang Jianfeng (nickname Xingdian) serving as the first dean, and investing 100 billion yuan in three years to explore cutting-edge technologies.

It advocates open collaboration between industry, academia and research institutes, and will conduct research and exploration of cutting-edge technologies such as artificial intelligence, the Internet of Things, and financial technology. The first batch of research areas announced mainly include: quantum computing, machine learning, basic algorithms, network security, visual computing, natural language processing, next-generation human-computer interaction, chip technology, sensor technology, and embedded systems.

It is reported that the establishment of the "Damo Academy" will also become a core component of Alibaba's "NASA Project" that it has always placed high hopes on.

As early as March last year, Alibaba launched the "NASA Project" for core technology research, saying that its goal for the next 20 years is to build the world's fifth largest economy, create 100 million jobs for the world, serve 2 billion people across borders, and create a profitable platform for 10 million companies.

The name "Damo Academy" is a transliteration of DAMO, and its full name is The Academy for Discovery, Adventure, Momentum and Outlook. The word "Damo" is also borrowed from the meaning of Damo in martial arts novels, which represents the highest level of martial arts cultivation.

In Jin Yong's martial arts novels, the Dharma Academy of Shaolin Temple is where Bodhidharma practiced Zen and martial arts. It is the highest-level martial arts research institution in Shaolin Temple. There are more than 30 monks of the same generation as Abbot Xuanci, and only 8 of them are qualified to enter the Dharma Academy.

The strongest martial arts and the deepest enlightenment are hard indicators; similarly, the "Damo Academy" established by Alibaba hopes to gather the world's top scientific forces, concentrate on researching core technologies for the future, and truly achieve "the greatest heroes benefit the country and the people."

At present, the official website of DAMO Academy has been officially launched. On it, you can see the latest and most comprehensive introduction to DAMO Academy, covering the five major research areas, two major cooperation ecosystems, and talent reserves and needs of DAMO Academy.

Among them, the Urban Brain Laboratory led by Hua Xiansheng, a Ph.D. in applied mathematics from Peking University and IEEE Fellow, also emerged and was fully displayed to practitioners.

Alibaba City Brain Lab is committed to opening up urban data pipelines, exploring data value, and building new urban infrastructure through the Internet and artificial intelligence.

It has been implemented in cities and countries such as Hangzhou, Suzhou, Shanghai, Quzhou, Macau, and Malaysia, covering areas such as transportation, safety, municipal construction, and urban planning. It is one of the first national artificial intelligence open innovation platforms and one of the largest artificial intelligence public systems in the world.

Its research directions include:

Multimodal big data perception: Achieve comprehensive and multidimensional perception of urban participants by integrating multi-source heterogeneous data through multi-perspective learning;

Urban traffic prediction and intervention : Based on the large-scale road network structure, analyze, predict and intelligently intervene in traffic congestion management issues;

Urban large-scale parallel heterogeneous computing : Accelerate the computing and processing process on massive real-time heterogeneous data networks through parallel heterogeneous computing;

Perception and understanding of complex urban environments : effectively model the perception of urban environments and design adaptive computer vision algorithms that are robust to the environment;

Urban visual search engine : Use the dynamic feature information of the video to model the characteristics of pedestrians and behaviors, and further complete search and identification;

Urban municipal planning and public resource analysis : Based on big data intelligent analysis and combined with the laws of urban development, intelligent analysis and decision-making are carried out on the layout of urban infrastructure and allocation of public resources;

The products and applications it covers include:

Tianji : Traffic and pedestrian flow prediction system. Through historical and real-time video data in the area, it accurately predicts the future traffic and pedestrian flow in the entire area in real time, provides reference for road diversion and control decisions, and avoids safety hazards such as congestion and trampling. At present, the accuracy rate of predicting traffic and pedestrian flow in the next 1 hour is over 90%;

Tianjing : Municipal construction and management system. It provides video automatic patrol and alarm services for municipal events to various government functional departments such as urban management, safety supervision, fire protection, housing and construction, and public security, assists manual inspections, eliminates hidden dangers in municipal construction, and improves the level of intelligent municipal management;

Tianying : A progressive video search engine. Based on real-time search of global video resources, it can quickly locate specific objects, such as finding missing persons and tracking hit-and-run vehicles. It only takes 1-2 seconds to process, and the pedestrian recognition accuracy rate reaches over 96%;

Tianyao : Full-time and full-area traffic automatic patrol alarm system. It can perceive traffic events and accidents in the city in real time, automatically detect abnormalities in all elements of people, vehicles, objects, and events, automatically identify traffic accidents and violations, and push them to the command center within 20 seconds, with an accuracy rate of more than 95%. It can realize 7*24 hours of uninterrupted inspection of urban traffic, reduce the workload of traffic police on road patrols, reduce the safety risks of traffic police, and free traffic police from the task of viewing monitoring, thereby improving law enforcement efficiency;

TianQing : A large-scale visual computing platform for the city brain. It includes three major components: video access system, real-time/offline computing system, and visual search system, providing a complete large-scale visual computing solution. "TianQing" has achieved rapid and flexible deployment in the cloud. It is an innovative product for security, providing customers with intelligent analysis capabilities on demand and effectively improving the efficiency of intelligent analysis. "TianQing" can achieve a thousand-fold acceleration of video analysis, processing 16 hours of video in just 1 minute.

Previously, in an exclusive interview with Leifeng.com, Hua Xiansheng said that the "Eye of the City" in the city brain is one of the important projects for commercialization.

There are probably hundreds of thousands of cameras deployed in first-tier cities. A single camera generates a huge amount of video data every day, but the value of this data is actually quite limited.

Therefore, Alibaba observes cars, people, and non-motor vehicles through the city's eyes, and uses visual computing to count the number of vehicles, models, license plates, lengths, speeds, driving paths, pedestrians, and other information. After collecting city data, an index will be established, which can be used in practical applications to find lost children and vehicles involved in accidents.

In the past, when visual technology was weak, only rough vehicle information could be obtained through cumbersome means: such as obtaining sampling data through GPS; burying ground sensor coils under the road and counting according to the pressure of the vehicle body. Both GPS and ground sensor coils can actually collect limited information about vehicles, and ground sensor coils often malfunction.

What the city brain needs to do is to collect data from the entire city and complete the understanding of urban accidents and events: know where the traffic jams are and where there are car accidents. After analysis, it can quickly issue instructions to control traffic lights and close intersections, and estimate the impact of accidents and events on traffic in the future.

During actual operation, the city brain uses real-time and full urban data resources to globally optimize urban public resources, instantly correct urban operation defects, and achieve triple breakthroughs in urban governance models, service models, and industrial development.

It can better enhance the government's management capabilities, solve prominent problems in urban governance, and realize intelligent, intensive and humanized urban governance; it can provide more accurate services to enterprises and individuals anytime and anywhere, save public resources, play a catalytic role in industrial development, and promote the transformation and upgrading of traditional industries.

At the same time, in the content launched this time, the City Brain also disclosed its overall architecture, which includes four major platforms, namely:

Application support platform : an open application support platform that prospers the industrial ecology and saves natural resources by consuming data resources;

Intelligent platform: An open intelligent platform that uses deep learning technology to mine gold mines in data resources and enable cities to have the ability to think;

Data resource platform: real-time aggregation of data from the entire network, making data a real resource, ensuring data security, improving data quality, and realizing data value through data scheduling;

Integrated computing platform: Provides sufficient computing power for the city brain with extreme flexibility, supports real-time computing of all city data, EB-level storage capacity, PB-level daily processing capacity, and real-time analysis of millions of videos.

It is worth noting that in this topology map, the algorithm partners of the intelligent platform saw Yitu's Dragonfly Eye and Megvii's Face++, while SenseTime, a technology company whose Series C investment was led by Alibaba, failed to participate in the construction architecture of the city brain. What do you think about this?

Alibaba DAMO Academy's official website is launched today, with 5 major research areas and 14 laboratories fully open to the public

Musk was sued by the US Securities and Exchange Commission, but he is not the American Jia Yueting

Bitmain, the world's largest mining machine company, has started its journey to go public. Will it be a new benchmark in the field of AI chips in the future?

Silicon Valley giants' "privacy legislation" hearing: not the trial of the century, but redemption