China is an agricultural country with a history of more than 5,000 years, but it is not an agricultural power. In recent years, smart agriculture has become an important industry in promoting the application of 5G, cloud computing, artificial intelligence and other technologies.
One of the distinguishing features of smart agriculture from traditional agriculture is that the industrial chain has been greatly extended, forming an industrial system that closely combines pre-production, in-production and post-production. The purpose is to combine smart thinking with information technology, computer technology and other advanced science and technology to achieve sustainable agricultural development. Let agriculture produce higher output, better quality, lower cost and less environmental pollution.
Artificial intelligence (AI) is an important driving force for the new round of scientific and technological revolution and industrial transformation. The application of AI in agricultural production mainly integrates computer and network technology, Internet of Things technology, 3S technology, wireless communication technology, audio and video technology and expert wisdom to realize intelligent management such as visual remote diagnosis, remote warning and remote control, improve the quality and efficiency of agricultural production, while improving the operator's sense of experience and comfort.
Flying OARS has gathered more than 2.3 million developers, serving 90,000 enterprises, and created more than 310,000 models based on the flying OARS platform. Flying OARS are playing a role in many fields related to the national economy and people's livelihood, such as city, industry, electricity, communication and so on, helping more and more industries to complete AI empowerment and realize industrial intelligent upgrading.
AI is used in the whole process of agricultural production, from prenatal to postpartum, and in various subdivided scenes. For example, PaddleDetection is a unified framework for object detection launched by Baidu Feizuo in the prenatal section. Support existing RCNN, SSD, YOLO and other series models, support ResNet, Resnet-VD, ResNeXt, resnext-VD, SENet, MobileNet, DarkNet and other backbone networks. For different business scenarios in the agricultural field, crop identification and pest detection (performance, target size, accuracy, etc.) can be selected from different module combinations in the framework to get the most suitable model and achieve the task.
Taking YOLOv3 model as an example, compared with the best similar products of the same period in the previous version, the training speed based on COCO data set exceeded 40%, and the mean Average Precision mAP of verification set was 38.9%, exceeding 1%. In this upgrade, the flying paddle engineers, in the spirit of craftsman of striving for excellence, further enhanced the model. The COCO data set mAP was up to 43.2%, and the training speed was also increased by 40%. Based on YOLOv3, a variety of complete model compression schemes were open sourced, bringing YOLOv3 to a new level!
In the production process, hydroponics is being promoted as a new environmentally friendly technology, as agriculture also seeks to transform itself. Before the introduction of AI capabilities, water distribution relied on agricultural professionals and experience, and quality judgment was made during the production process. Once personnel were negligent, 15% to 20% of the production was wasted. After the introduction of EasyDL for model development, automatic machine identification was realized, real-time management of vegetable growth was completed, which helped to reduce the output of defective products, improve product quality, increase output by 10%-15%, and reduce the cost of production materials such as seeds, substrates and nutrient solution by 10%-15%. The productivity of agronomists increased threefold.
Beyond that, how can AI capabilities empower the agricultural landscape, improve productivity and reduce costs elsewhere?
On October 29, China Agricultural Machinery Industry Association and Baidu Fei Zuo invited Baidu's outstanding architects and product owners to live analyze the practical experience of AI in agricultural production. The cases covered all links of the whole industrial chain, such as cultivated land identification, bug detection, automatic crop sorting, and import quantity detection. There are lots of dry goods, scan the code to watch, be there or be there!
