AI accelerated "video / visual" development, recently, Shenzhen will usher in one of the areas of artificial intelligence authority tournament - the first "National Competition Artificial Intelligence" (The First National Artificial Intelligence Challenge). The contest is organized by the Shenzhen Municipal People's Government and the Shenzhen Municipal Bureau Commission, Ministry of Science Pengcheng laboratory and guide the establishment of a new generation of artificial intelligence industry technology innovation strategic alliance (AITISA) co-host, and join hands Artesyn Technologies, Li Fei sky, peace Technology, Skyworth Group co.
Which is one of the lab laboratory Pengcheng Guangdong Province, the first batch of four provinces, the network is committed to undertake an important strategic task information field, to play a leading role in promoting the development of discipline and resolve major national scientific and technological problems.
Competition nineteen full implementation of the spirit of "a new generation of artificial intelligence planning", and hopes about the "AI + 4K HDR" and "heavy pedestrian recognition" technology bottlenecks are two practical problems and solutions, pulling people together to promote national AI "video / by leaps and bounds in the field of vision "docking leading industry resources to promote industrial innovation elements, academic, capital, personnel and other integration and development.
Enabling policies, AI innovations
As a science and technology, AI has always been inseparable from the development of events to promote the important challenges of the world. There ImageNet before large-scale visual recognition Challenge has been called a milestone in the history of AI; the "Father of neural networks" Geoffry Hinton and his team have been included into the contest NIPS paper is considered deep learning hot to open; after American unmanned DARPA 2004 challenge achievements of today autopilot lidar manufacturer Velodyne.
The upcoming start of this month in Shenzhen, the first "National Competition artificial intelligence" is not just a challenge for the world event; also the "CPC Central Committee and State Council on the support of the Shenzhen building socialism with Chinese characteristics first demonstration zone of opinions" after the introduction, Shenzhen hosted the first field of AI authoritative professional events.
Therefore, competition from the proposition either design, data preparation, or evaluation methods are fully reflected the strength of the country's top AI laboratory and industry giants such as Tencent, designed to stimulate innovation and vitality of the creative potential of the whole society, to further accelerate our high level artificial intelligence personnel and troops of mining, culture and growth.
The contest problems forefront of the industry focus
"Opinions" clearly states that support the construction of Shenzhen 5G, artificial intelligence, a network of space science and technology, life and biomedical laboratory information and other major innovation carrier. The contest will also be the title match precise focus AI, 4K, 5G and other frontier areas, set up a "AI + 4K HDR" and "heavy pedestrian recognition" two contest entries, while the urgent need for a breakthrough in the two competition entries are current industry technical bottlenecks.
This is not only for nineteen spirit of "a new generation of artificial intelligence planning" are fully implemented, completely fit the needs of the times and the practical problems of integration of AI, but also on behalf of the Shenzhen area first to lead the country to open up the field of AI industry, academia, research path, the world AI determined to train young talents. At the same time, but also for the development of Guangdong, Hong Kong China, especially large Bay Area Artificial Intelligence industry, providing important support and impetus to further promote the outstanding quality of artificial intelligence and artificial intelligence personnel projects landing in Shenzhen.
Moment, transformation and development of emerging technologies artificial intelligence, ultra-high definition, cloud computing, big data, 5G and so is having a profound impact on the media industry. In the ultra-high-definition content production processes, it is prone to various problems, such as: quality degradation, low resolution, brightness, gamma change, change color space, the synthesis of low performance. "Video / visual" field is the current main application areas of artificial intelligence, "AI + 4K HDR" tournament entry represents the direction of the depth of integration of science and technology and culture, will bring video quality leap, greatly improving the quality of life of people.
The "heavy pedestrian recognition" Race entry is the focus of academic research and industry, and an important breakthrough in the next "face recognition" of. In practice, when there is no overlap between the shooting range of the camera, since there is no continuous information retrieval difficulty also increases a lot. Thus there are many technical issues need to be resolved artificial intelligence expert.
I believe that with the implementation of events, "AI + 4K HDR" and "heavy pedestrian recognition," a series of practical issues will be the country's top experts in AI attention and work together to solve, to further promote leapfrog development in the field of AI.
Talent Addition
Of course, in the Government's strong support, in addition finalists will share the huge prize pool (before each tournament prize items are intended to reward one million yuan, 50 million, 20 million, there are a number of award winning, ten thousand yuan Tencent cloud vouchers, etc.), players have the opportunity to race before the single item of the top 20 teams get Pengcheng laboratory, Artesyn Technologies, Skyworth Group, Yun Tian Li Fei, Ping An Technology recruitment green channel, government research funding, personnel policy, research and development space, such as full support.
We interviewed this important tournament organizer - Pengcheng laboratory, they focus on the process responsible for the overall competition, including: competition program, competition rules, competition affairs management.
Pengcheng laboratory staff involved in the preparations for the contest mentioned, this contest is an important measure to build Shenzhen first demonstration area, promote innovation and development as well as to the introduction of the national talent; competition is intended to be based on an international perspective, the Shenzhen into an international advanced AI industry a new generation of strategic high ground, to create the next AI field a unicorn. He also revealed, this competition also prepared online expert technical guidance to participants, online video training, basic algorithm / code hosting, investment institutions docking, docking incubators and other services and support.
Related personnel, said, "National Artificial Intelligence Competition" future potential to become China's largest artificial intelligence official events. While stressing the contestants, contest entries set two match involving a number of AI technology, "Getting easy, but difficult to achieve top-leading and comprehensive." Reminded the players were able to increase exchange and communication within the industry, the use of the advantages the complementary combination of competition.
Artificial intelligence is a new round of technological revolution and industrial revolution of the important driving force. The AI contest for most AI developers, that would be great personal development opportunities not to be missed! As a strategic media partner of the event, will contest Lei Feng network first-hand data compilation edited as follows.
Important events
l Time: October 2019 -2020 February
l Location: Shenzhen, China
l Organizers: Shenzhen Municipal People's Government
L Competition Organizing Committee Director: Chen Rugui Gawain
L Competition Director Technical Committee: Academician Ding Wenhua
Introduction tournament title
The events were set up two major events, namely the "AI + 4K HDR" and "heavy pedestrian recognition," race around the term "to address the challenges in the field of 'video / visual processing' by artificial intelligence" combined with the most hope outstanding talent and strength of AI algorithms, and achieved a breakthrough leap forward in these areas. The title match details are as follows:
A game item: AI + 4K HDR contest
"AI + 4K HDR" competition entries focused on artificial intelligence applied to 4K video processing; in 5G network, one of the most important applications is 4K / 8K video. The first half of 2019, the Ministry of Industry, radio and television, CCTV made it clear in 2022 4K / 8K users reached 200 million, 4K / 8K industrial scale up to 4 trillion yuan target.
Technical Difficulties
4K Ultra HD content shortage has been the industry's largest short board and difficulty. Currently on the market can really meet the 4K resolution (/ HDR / WCG / HFR / 10bit) and other multi-dimensional indicators of ultra-high-definition video is quite small, but also for the old sources, the video section above meet the target of a lack of intelligence tools and methods enhanced optimization. Among them, especially the SDR to HDR conversion and image quality enhancement is a lot of technical difficulties.
Breakthrough goal
AI may become breaking method and a useful complement to 4K / 8K video content. Wherein, the AI algorithm not only can enhance the low mass for high-quality video sources, plus noise and partial repair / video damaged (removing blockiness compression / noise, etc.); Resolution and lifting (four times over points), frame rate up the effect is far superior to the traditional methods, the SDR conversions HDR, image bit depth increase, gamut expansion, denoising may be done even better than the original captured equivalent level sources.
In order to better study the problem, build a industry's largest, most extensive reconstruction of 4K HDR video data sets, covering different difficulty, noise models under different scenarios, it is very crucial and important.
Reference material
Super-sub publicly Code links:
l https://github.com/thstkdgus35/EDSR-PyTorch
l https://github.com/xinntao/EDVR
SDR2HDR open Code links:
l https://github.com/gabrieleilertsen/hdrcnn
Race Item two: heavy pedestrian recognition contest
Identifying heavy pedestrian, the pedestrian is also called re-identification, i.e. re-recognition of the pedestrian. It is the process of establishing a correspondence between different images of the pedestrian is not the field of view of cameras with overlapping coverage.
Technical Difficulties
When there is no overlap between the shooting range of the camera, since there is no continuous information retrieval difficulty increases. Currently, the bottleneck identification heavy pedestrian main face there are many, for example: it is difficult to identity data annotation pedestrians, pedestrian image susceptible apparent illumination, angle of view, posture, occlusion factors, leading to the identification of low accuracy.
Breakthrough goal
As a research topic of common concern academia and industry, and heavy pedestrian goal is to identify a large number of surveillance video from accurately identify the target staff queries. Thus, heavy pedestrian recognition is carried out in the cross-emphasized video camera in search of a particular pedestrian, to make up for the shortcomings of existing face recognition systems, and machine learning mode by the depth from the human back, side or face under conditions of ambiguity, a person will act to restore the track to achieve the target figure of identification, retrieval and tracking.
Reference material
DukeMTMC-reID offers 23 kinds of attribute data marked DukeMTMC-attribute for download:
l https://github.com/vana77/DukeMTMC-attribute
Disclosed in heavy pedestrian identification code linking:
l CNN+Pose：https://github.com/erichuang0771/PoseBox-Reid
l Baseline：https://github.com/zhunzhong07/IDE-baseline-Market-1501
l GAN + heavy pedestrian identification: https: //github.com/layumi/Person-reID_GAN
Important time node
l 2019 Nian 10 - November: preliminary round, the contestants based on the training set, validation set, preliminary test set, the algorithm design, model training and evaluation, and submit the results to predict the race;
l 2019 Nian 12 months: semi-final stage, opening up a rematch test set to enter into the semi-finals of the players to submit the results of scores, rankings, code verification, individual game players will also examine the operational efficiency of the algorithm;
Carry out the final stage, the top 10 players into the final quarter-final match items held in Shenzhen, the final use of on-site technical competition, field defense, etc.: l January 2020.
bonus system
Contest were set up two types of bonuses (Shenzhen provided), giving the winners rewarded.
Final awards (each track)
Enter the contest finalists and won the first prize, second prize, third prize-winning team, were obtained 1 million, 500,000, 200,000 in prize money. Awards detailed settings as follows:
Winner Awards
If you enter the contest finalists, but did not award-winning team, will race side also awarded Excellence Award (tentative four teams), and the $ 20,000 prize money each team will receive.
Registration Related
Competition is open the whole of society, universities, enterprises, research institutes and individual corporate teams, meet the conditions set forth in the official website are invited to participate in the competition. Entry section conditions were as follows:
l teams consisting of at least one person and the same track each person can only participate in a team;
l Team members registration information must be true and accurate and effective;
l participating teams do not limit nationality and age, but each team has at least one Chinese nationality players.
Current enrollment channel is turned on, just on the line test, there are many players can not wait to sign up, which can also be seen fanatical attractive contest for AI people; at the same time, Pengcheng laboratory has also put up a booth contest at CCF preheat booth card, attracted many exhibitors gathered to watch. It is foreseeable that the first "National Artificial Intelligence Competition" will certainly be gathered in the AI experts, technology and passion to a burst Sheng war.
Registration Address: http: //www.china-ai.ac.cn/
(↑ scan code may also enter the registration page ↑)
In addition, on October 22 (Zhouer), Pengcheng Lab will hold a grand launching ceremony of the first "National Artificial Intelligence Competition", the ceremony more event details and other benefits will be announced one by one.
Launching ceremony broadcast address: https: //www.yanxishe.com/development/69