Krishi-Vision: A Multimodal Bengali Dataset for Crop Disease Detection and Remediation

Overview:
Krishi-Vision is a comprehensive multimodal dataset designed to advance AI-driven agriculture in Bangladesh. It contains 998 high-quality images of various crop diseases paired with detailed expert-level diagnostic text in Bengali. The dataset follows a structured "Instruction-Response" format, making it ideal for training both Image Classifiers and Large Language Models (LLMs).

Importance:
Agriculture is the backbone of Bangladesh's economy, yet farmers often lack immediate access to agricultural experts. This dataset bridges that gap by providing a digital foundation for building AI assistants that can identify diseases from images and provide actionable remedies in the native language (Bengali).

Use Cases:
a) Visual Question Answering (VQA): Training multimodal models (like LLaVA or CLIP) to understand agricultural context.
b) Crop Disease Classification: Developing AI apps for real-time disease identification.
c) Agricultural Chatbots: Building chatbots that provide precise remedies in Bengali.

Data Source & Credit:
The raw data has been sourced from "Krishoker Janala" (https://krishokerjanala.gov.bd), a public service portal by the Government of the People's Republic of Bangladesh. We acknowledge and credit the original creators for their contribution to the farming community.