On April 20th, Baidu's "Sasahara Plan 2018" and Baidu AI Developer Combat Quarter 2 (first station in Beijing) released three major messages: Open EasyDL platform, publish "deep learning engineer evaluation standard", and face recognition interface For SMEs and developers will be permanently free.
In last year's "Sasahara Plan 2017", Baidu once introduced the open technology capabilities of Baidu AI to developers across the country. This year, they will focus more on "helping partners from all walks of life to really use Baidu AI."
First, open EasyDL platform
Let AI Xiaobai also be able to train a customized AI model at a zero threshold
In order to reduce the entry threshold for artificial intelligence, Baidu has fully opened its customized training and service platform.EasyDL. Xie Yongkang, technical manager of Baidu's AI Technology Ecology Department, introduced that EasyDL is a customized model training and service platform. Users can perform simple page operations based on text prompts. It takes only four steps to train business-customized deep learning in 10 minutes. model. This means that even if you do not know the technical whiteboard of code programming at all, you can start quickly with zero barriers.
Xie Yongkang said that EasyDL has four characteristics: ready-to-use, lighter, higher-precision, and stronger security, and is extremely friendly to newly-entry AI developers. Currently, EasyDL has opened up image classification, object detection and customization training. In the future, it will gradually open up its ability to customize voice recognition, as well as multi-environment deployment and data services.
In addition, Lei Fengnet learned that the EasyDL “Unbelievable Contest” will also officially start on May 1.
Second, empower AI three initiatives
Provides guidance for growth learning engineers
Baidu also announced a number of new initiatives that have enabled the development of deep learning, bringing deep learning capabilities to all walks of life.
First, Baidu announced its opening upDeep Learning Engineer Evaluation Standard, which provides reference standards for 16 categories of competency items, including general capabilities, professional knowledge, professional capabilities, industry knowledge, and organizational management, as well as 16 dimensional competency-based reference standards, to provide deep learning engineers with "universal competence" to "professional Skills and then the "organization management" capability model pyramid, to guide developers to grow.
Secondly, Baidu has built a complete system of deep learning content. Through on-line and off-line learning courses, it supports deep-learning talent development. Whether it is entry-level players or seeking advanced promotion, Baidu can find a suitable development for itself. path. Baidu also co-hosted with Beijing University of Aeronautics and Astronautics under the guidance of the Ministry of Education and the Ministry of Industry and Information Technology. The "Information Technology New Engineering, Industry, Research and Development Alliance Artificial Intelligence Cooperative Education Committee" and the "Teacher Training Working Committee" jointly organized a series of in-depth learning teachers. The training class tailored training for young teachers in universities to help reserve talents in universities.
Third, Baidu accelerated the pace of deep learning from open source to landing in the form of commercial operations. At the end of last year, Baidu and CodaPaddlePaddle AI Contest, Through the competition model to help solve real problems in the industry. A total of 496 teams participated in the competition and 281 teams received 909 high quality works. In the course of the conference, Yu Youping, general manager of Baidu's AI technology ecology department, awarded prizes of 50,000, 20,000 and 3,000 to the winners of the first, second and third prize respectively.
Third, permanent free
SMEs are totally free, and the cost of calling companies with large demand is reduced by 40%
Cost is a barrier to the access of developers or companies to artificial intelligence. Yu Youping first emphasized again that Baidu AI open platform voice technology is “permanently free”:
Subsequently, he also announced the adjustment of face recognition commercial interface charging mode:
At present, Baidu face recognition interface fully adopts the QPS monthly subscription model. It gives free developers 2QPS and 5QPS monthly configuration amounts to developers and certification companies. It only needs to pay more than 500 yuan/month QPS or 50 yuan/day. /QPS. After this adjustment, the free allocation of certified companies increased to 10QPS/month, while the price exceeding the free portion also fell to 300 yuan/month QPS or 30 yuan/day/QPS. This means that SMEs with low QPS for face use can still be used for free forever, and for enterprises with a large demand, the call cost for face recognition is reduced by 40%.
Yu Youping said that Baidu’s downward adjustment of face charging standards is to hope that the technology will be “popularized” as soon as possible, so that all partners can use Baidu’s AI technology to make more products and innovations and benefit the whole society.