CropChain: An AI-Powered Agricultural Social Platform Integrating Rice Disease Detection, Expert Collaboration, and Digital Marketplace Services for Smallholder Farmers

Rice farming supports millions of smallholders in Pakistan, yet these growers face several problems at once.Diseases are often noticed only after they spread, trained agronomists are hard to reach in rural areas, and produce is usually sold through intermediaries who take a large share of the margin.Most digital tools tackle only one part of this chain.We present CropChain, a single application that brings four functions together: automated diagnosis of rice-leaf diseases, an Urdu advisory chatbot based on a large language model (LLM), a community feed for expert and peer collaboration, and a direct marketplace.The application runs on Android and has a companion web client.The diagnosis engine was trained by transfer learning on a public dataset of 2,628 rice-leaf images in six classes.We fine-tuned three pre-trained backbones as feature extractors and compared them under the same settings.EfficientNet-B0 gave the best validation accuracy at 95.08%, followed by MobileNetV2 (93.56%) and ResNet-50 (89.39%), and was chosen for deployment.Because these figures come from curated validation images, they are best treated as an upper bound on performance in the field.We also assessed user acceptance with a 54item questionnaire completed by 52 respondents from Pakistan's rice belt.Over the 43 evaluative Likert items the mean rating was 4.35 of 5.00, with 89.1% agreement.The unified-ecosystem items scored highest (mean 4.44), and 94.2% of respondents agreed that combining AI, expert, and community input leads to better decisions than AI on its own.Overall, the findings suggest that farmers place more trust in an accurate classifier when it is embedded in a human-in-the-loop platform than when it stands alone, and that the integrated design is associated with easier access to agricultural services.