The "Beyond Visible Spectrum: AI for Sustainable Agriculture" challenge unites cutting-edge hyperspectral imaging with advanced AI techniques to revolutionize precision farming. This initiative invites researchers, industry experts, and students from around the globe to develop innovative solutions that address the pressing issues of crop disease detection and data scarcity in agriculture. There are two main tasks in the challenge. 

# Hyperspectral Data Analysis for Crop Disease Classification

The first task focuses on hyperspectral data analysis for crop disease classification. Participants are tasked with designing deep learning models that accurately distinguish between diseased and healthy crop areas using real hyperspectral imagery. This approach leverages the unique capability of hyperspectral sensors—which capture detailed spectral information beyond the visible spectrum—to detect subtle physiological changes in crops. By enabling early and precise disease detection, the task aims to reduce yield losses and mitigate the economic impacts of crop diseases.

# Exploring Synthetic Hyperspectral Data Generation

The second task centers on synthetic hyperspectral data generation. Here, participants explore the potential of generative models such as GANs, VAEs, and diffusion models to create high-fidelity synthetic datasets that mirror the spectral characteristics of real agricultural imagery. This synthetic data serves as a critical augmentation tool, helping to overcome the challenges posed by limited annotated datasets and enhancing the robustness and generalization capabilities of AI models. The ability to simulate a range of crop conditions, including varying stages of disease progression, is expected to foster the development of more resilient and scalable AI-driven solutions.

The challenge is built on the success of previous competitions, which have shown growing participation and interest.  In last year, a total of 190 people from more than 10 countries in Europe, Asia and North America entered the competition by the deadline date, with a total of 126 submissions of results.

Participants will work with comprehensive, pre-processed hyperspectral remote sensing datasets obtained via UAV-mounted sensors covering wavelengths from 450 to 950 nm. These datasets, comprising high-resolution patches with rich spectral detail, provide a realistic testing ground for innovative techniques aimed at addressing modern agricultural challenges.

Ultimately, the "Beyond Visible Spectrum" challenge is more than just a competition—it is a dynamic forum for interdisciplinary collaboration that bridges the gap between state-of-the-art AI research and real-world agricultural applications. By driving advancements in crop disease monitoring and synthetic data generation, the initiative aspires to contribute significantly to global food security and sustainable farming practices, paving the way for a smarter, more resilient future in agriculture.

# Workshop/conference
This challenge is part of the conference on "IEEE International Conference on Image Processing, IEEE ICIP 2025".