Smart Swine Growing – SME SEEDx Development Challenge 2020
In this blog, you will learn about a solution that has been developed for the SME SEEDx Development Challenge 2020.
In year of 2019, an infectious disease called “African swine fever” swept across Asia. Hundreds of millions of pigs were infected and killed, and the pig industry suffered heavy losses. Pig farms located in most areas of China have been closed for nearly 6 months (farmers can only go out and cannot enter the farm), and the pig farms are seriously understaffed and the staff are becoming over occupied. In order to reduce the work intensity of frontline workers inside the pig farm and reduce the spread of disease caused by human intervention in the breeding process, we set the goal of using AI technology to automatically collect data throughout the breeding process.
Solution Demo: Smart Swine Growing
Solution Name: Smart Swine Growing
Use the camera to shoot and monitor the behavior of pigs in the barn. The system uses AI-deep machine learning technology to achieve automatic inventory and weight estimation of biological assets, thus to reduce manual inventory and intervention.
Solution Use Case
Real Customer Use-case:
This is a real customer case, it’s a Chinese state-owned enterprise. Due to the impact of African swine fever and the increase in labor costs, it is necessary to establish a fully automatic intelligent pig farm, to achieve automated collection of production data during pig farming. This will save a huge amount of labor cost and guarantee biosafety and food safety.
Our customers are pig companies and farms, and those companies and the whole industry will benefit from this solution.
- Discovery of abnormal pig status:
The abnormal pigs (such as diseases) in the farms are usually detected by experienced technicians through manual patrols. However, in consideration of biosafety (African swine fever control), veterinarians need to reduce the frequency of human-to-pig contact (without special circumstances, it is better not to enter pig pens). This leads to a contradiction between production performance and disease prevention. Any delay of patrol would introduce additional production losses.
- Inventory and weight valuation of pigs in the process of transfer:
Each time the farm transfers pigs, it is necessary to count the number of pigs transferred out and select some pigs for weighing (to evaluate the average individual weight of pigs currently transferred out). The above two data (quantity & individual weight) are not only the summary of the production performance of the upstream production workshop, but also the beginning of the downstream production workshop. These are very important data. However, these two data are often inaccurate in the actual production process, as the numbers in the production workshop do not always match, which leads to inconsistency between upstream and downstream production workshops, and it is more difficult to evaluate the production performance of the upstream and downstream production workshops.
Target Market: Large pig farms (more than 300,000 pigs per year) and Large breeding companies (more than 1.5 million pigs per year).
Deploy cameras on the farm to capture live video of the herd and pigs, and upload the video to Ali Cloud’s OSS service. Ali Cloud’s PAI platform can read videos (photos) on OSS and cooperate with Smart Swine Growing’s optimized algorithm model to achieve real-time biological asset inventory and weight estimation.
Aliyun PAI（Platform of Artificial Intelligence）
Machine Learning（Deep Learning）
We will continue to improve existing features and develop new scenarios. The goal is to automate the data collection of the entire breeding production process within 3-5 years. The ultimate goal is to achieve unmanned breeding in the pig farm.
Partner: MTC (Partner of SAP S/4 HANA / SAP S/4 HANA Cloud/ SAP Business byDesign/ SAP Business One/ SAP SuccessFactors)
Team Name:AG business unit
Team Members: Robin Xu,Terry He,Lee
Conclusion: There are still few connection between AI and the actual pig production process. The factors that cause this problem are mainly the low informationization level of pig enterprises. At the same time, the current AI algorithm still cannot be directly applied in the scene of pig farms. It is necessary for us to continuously iterate the algorithm to be close to pig production, and it is necessary to combine multiple technical methods to create simple and practical low-cost technical solutions in the future.
For more information about the development challenge, you may refer to the SME SEEDx Development Challenge 2020.