Page 854 - AI for Good Innovate for Impact
P. 854

AI for Good Innovate for Impact



                          Use Case 4: Divine Farmer Agricultural AI Model

















                      Country: China

                      Organization: China Telecom Wanwei Information Technology Co., Ltd.

                      Contact Person: BinBin Wen,  173735562@qq.com,

                      ShaoJie Zhang, 18919821844@189.cn


                      1      Use Case Summary Table

                       Item              Details

                       Category          Smart Agriculture
                       Problem           Over reliance on individual experience, shortage of technical personnel,
                       Addressed         difficulties in promoting advanced agricultural technologies, low agricul-
                                         tural informatization level, limited service coverage, inconvenient access to
                                         technical information.

                       Key  Aspects  of  Artificial Intelligence (AI) - powered agricultural Question & Answer assistant
                       Solution          that leverages Interactive Natural Language Processing (NLP) for real-time
                                         planting guidance and knowledge dissemination. It also includes a pest and
                                         disease identification module using image-based AI recognition with over
                                         86% accuracy, offering tailored treatment plans. Additionally, the platform
                                         integrates price forecasting capabilities through NLP and time-series anal-
                                         ysis to deliver reliable market trend predictions.
                       Technology        AI/ML, LLM, Computer Vision, Natural Language Processing (NLP), Data
                       Keywords          Analytics.
                       Data Availability  Private Data (e.g., proprietary agricultural production records, pest/disease
                                         image databases from partner farms, historical market price datasets).
                                         Sources not publicly disclosed.

                       Metadata (Type of  Text (for Q&A, price analysis input/output), Visual (for pest/disease identifi-
                       Data)             cation images), Structured Data (for price analysis  and  prediction).

                       Model Training  CV Models: Pre-trained YOLO fine-tuned on pest/disease datasets. Large
                       and Fine-Tuning   Language Model (LLM): Pre-trained Qwen2.5-32B for Q&A and price trend
                                         analysis. Time-Series Forecasting: LSTM models for market predictions.

                       Testbeds  or  Pilot  AI - powered agricultural assistant offering technical guidance, price
                       Deployments       forecasting, and decision support to farmers through natural language
                                         interaction [4]








                  818
   849   850   851   852   853   854   855   856   857   858   859