laitimes

Smart garbage recycling, pharmacies automatically grab medicines... Baidu Flying Propeller inspires AI hardware and software innovation

Reading Chuang / Shenzhen Business Daily reporter Chen Shu

On October 28th, Feipao China Tour - AI Empowerment Software and Hardware Product Innovation Special Session landed in Shenzhen, Baidu technical experts and industry experts focused on the theme of "Software and Hardware Product Innovation", focused on the actual situation in the development of Shenzhen enterprises and industrial clusters, directly attacked the key pain points in the development of intelligent software and hardware, and displayed Baidu's AI technology, application cases and service resources on the spot.

Smart garbage recycling, pharmacies automatically grab medicines... Baidu Flying Propeller inspires AI hardware and software innovation

Senior manager of Baidu's AI product business unit said in his speech that deep learning is the core technology field of artificial intelligence. As China's first industry-grade deep learning platform with independent research and development, rich functions, open source and open source, Baidu Feipao will comprehensively escort enterprises to achieve rapid AI transformation and upgrading from three aspects: multi-level technology products, industrial AI talent training and strong ecological resource support. At present, Baidu Feipao has gathered 3.7 million developers, empowered more than 140,000 enterprises, and lowered the threshold for AI development and application.

In the face of the increasingly high proportion of customized AI demand, the high cost and long cycle of traditional custom models, and the inability to accurately predict the project effect in the early stage, Baidu launched the Flying Propeller EasyDL Zero Threshold AI Development Platform, which enables application developers with zero algorithm foundation to obtain customized AI services as soon as 15 minutes.

In terms of edge deployment, EasyDL has built-in EasyEdge terminal and edge AI services, which can automatically complete the adaptation and acceleration of the model to the hardware, reduce the installation and deployment time by an average of 90%, increase the inference speed by up to 10 times, and generate the end-edge computing service in as fast as 2 minutes. Faced with the challenges of algorithm training, hardware selection, secondary development, and integration, Baidu simultaneously launched the EdgeBoard embedded AI solution, launched the capability-based hardware and software integration and EasyDL software and hardware integration solution, from model construction to hardware selection and adaptation, to application integration, providing full-stack AI technology services.

EdgeBorad has been widely used in intelligent hardware, video analysis and other fields, the conference site showed intelligent fruit and vegetable scales, intelligent garbage recycling bins, intelligent inspection drones, unmanned line patrol cars, robot dogs and other intelligent products and solutions based on EdgeBorad.

Guigu Intelligence, an enterprise engaged in the design and development of intelligent environmental protection products in Shenzhen, based on the EasyDL zero-threshold AI development platform, trained the available beverage bottle intelligent identification model within one day, and efficiently deployed it on the beverage bottle recycling machine and other intelligent device terminals through EasyEdge. At present, the smart beverage bottle recycling machine has been placed in a large number of shopping malls and supermarkets in Hong Kong, and more than 1 million beverage bottles have been recycled in more than two months.

In the face of the current pain points that the self-service dispenser in the hospital cannot accurately identify drugs, Beijing Zhuoyinda Technology Co., Ltd. Feipao EasyDL platform cooperation, in 24 hours to master the entire set of training processes from submitting pictures, intelligent labeling, model training, joint debugging and use, etc., the number of samples per label is reduced by nearly 85%, and the training time is reduced by 80% under the premise of ensuring the recognition rate, which significantly improves work efficiency. Today, Zoinda's intelligent unmanned medicine cabinet solution has been applied to the ground, reducing the pressure on medical staff and reducing the operating costs of selling drugs.

Up to now, EasyDL has been widely implemented in industrial manufacturing, safety production, retail FMCG, intelligent hardware, culture and education, government affairs, transportation and logistics, Internet and other fields, and continues to maintain the first share of China's machine learning platform in the IDC survey report.

Review: Tan Lugang

.AI

Read on