{ "fetched_at": "2026-10-19T07:11:45Z", "source_url": "https://www.kaggle.com/competitions/beyond-visible-spectrum-ai-for-agriculture-2024p2/data", "pages": [ { "name": "data-description", "content": "## Un-Supervised training data\n\nThe data for SSL training can be download [https://cropdis.s3.eu-west-2.amazonaws.com/s2a.tar.gz](url)\n\n## Supervised training data\n\n`kaggle competitions download -c beyond-visible-spectrum-ai-for-agriculture-2023-p2`\n\n##Submission Columns\n\n* `Id` - file name\n* `Category` - Aphid, Rust, RPH, Blast", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "# Introduction\n\nParticipants from all over the world can take part in the challenge. \nYou must be over 18, or of legal age in your jurisdiction, in order to participate.\n\n\n# Timing and duration\nThe Crop disease challenge will take place online from **20 April to 20 May2024**.\n\n\n# Creating and submitting a solution\n\n\n\n1. For the challenge, you will have access to the training datasets and a range of tools in the form of a Jupyter notebook (Starter Pack). No other datasets can be used for the development of your solution. However, pre-trained models are allowed.\n\n- The data for this competition is available in the Kaggle webpage. It can be downloaded or processed using Kaggle's computing services.\n\n- Here is a short code example for data reading and processing the data ([Starter](https://www.kaggle.com/code/robeson/starter)).\n\n2. Each participant will be allowed to upload their solutions every 12 hours (to avoid overload and unfair behaviors).\n\n3. The final submission of your solution must be made by **20 May, 2025**, and each valid submission must consist of a Jupyter notebook with the implementation of their final algorithm to allow the Organizers to reproduce the predictions.\n\n# Challenge evaluation\nThe evaluation of the solutions submitted will be performed in two stages:\n\n1. Quantitative analysis: live scoring of the quality of your solution throughout the challenge, which will be visible on the leaderboard on this platform; \n\n2. Qualitative analysis of your code performed by the Organizers. The qualitative analysis is performed only for proving the validity and reproducibility of your solution.\n\nThe final score is based solely on the quantitative evaluation. However, if the solution turns out to be not reproducible, or if any concerns about its validity arise, the organizers reserve the right to contact the team leader of the corresponding team to resolve such issues. If the issues cannot be resolved, the team will not be eligible to receive prizes.\n\nFor more details on the metrics, see the Metrics section in the Data page.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "#Important Information\n\n\nPlease contact me (robeson1010@gmail.com) to verify the your results on our private dataset. We will award and certification after evaluation at August 18, 2024.\n\n@Vietnamese\n@Nguyen Hoang Anh 2903\n@ElyyLong Trương\n@In Processing\n@Long Trương\n@ElonMusk113\n\n\n# Overview\n\nThe International Conference on Pattern Recognition (ICPR) 2024 Competition on “**Beyond Visible Spectrum: AI for Agriculture**” presents an exciting opportunity for researchers and practitioners to advance computer vision techniques in agricultural crop disease monitoring. By leveraging the vast multi/hyper spectral remote sensing image datasets, participants are encouraged to develop innovative deep learning algorithms. The primary aim is to enhance the accuracy and efficiency of identifying crop diseases, thereby contributing to sustainable agricultural practices and global food security.\n\nParticipants in this challenge will have the chance to contribute to significant advancements in precision farming and crop monitoring techniques. By improving the deep learning vision models, the competition aims to drive forward the capabilities of agricultural technology, making it more efficient, accurate, and sustainable. The ICPR 2024 Competition on “Beyond Visible Spectrum” is not just a competition; it is a call to action for the computer research community to bring about transformative changes in how agricultural data is analyzed and utilized, paving the way for smarter, more sustainable farming practices that can benefit communities worldwide.\n\n\n\nIn this competition, participants are challenged to employ self-supervised learning techniques to identify various types of crop diseases from remote sensing imagery. This task addresses the issue of limited labelled data in remote agricultural regions and encourages the development of models capable of learning meaningful representations without extensive annotated datasets. \n\nSelf-supervised learning, at its core, revolves around an auxiliary task, often referred to as a pretext task, which utilizes auto-generated labels to guide the initialization of model parameters. These parameters are subsequently employed to enhance performance on downstream tasks such as classification, segmentation, and object detection. In essence, self-supervised learning entails two primary stages: the pretext task and real downstream tasks. The pretext task involves training a network in a self-supervised manner, leveraging auto-generated labels derived from unlabelled data to initialize model parameters effectively. By optimizing this objective through carefully designed pretext tasks, the network acquires the capability to extract meaningful representations from the input data. Subsequently, the trained network can be seamlessly transferred to supervised downstream tasks for real-world applications (e.g. classification tasks), where it can further enhance performance based on the learned representations [1-2].\n\nFor this challenge, we will provide a substantial remote sensing datasets covering arable land areas, which includes multi-spectral and synthetic aperture radar data (Sentinel 1/2). This dataset contains 500,000 images for participants to train their self-supervised models. Then a crop disease classification task and datasets will be established, covering wheat stripe rust, aphids, and rice blast and planthoppers. Participants use this data to fine turn their self-supervised models. The competition determines the performance of a self-supervised model by evaluating how much its performance improves over a model that does not use the self-supervised training dataset.\n\n\n[1. A generic Self-Supervised Learning (SSL) framework for representation learning from spectral–spatial features of unlabeled remote sensing imagery.](https://doi.org/10.3390/rs15215238)\n\n[2. The self-supervised spectral–spatial vision transformer network for accurate prediction of wheat nitrogen status from UAV imagery.](https://doi.org/10.3390/rs14061400)\n\n\n\n\n\n# Competition Schedule\n \n**Launch Date:** April 15, 2024\n\n**Submission Opening:** April 15, 2024\n\n**Submission Deadline:** July 21, 2024\n\n**Submission of Competition Reports Deadline:** August 18, 2024\n\n# Awards\n\nAll submissions before the deadline will receive a participation certificate\n\nThe top three places of each task will be awarded prizes to the value of:\n\n\n**1st Place:**\t$300\n\n**2nd Place:**\t$200\n\n**3rd Place:**\t$100\n\n\n\n\n# Challenge keywords\n\nRemote sensing; Hyperspectral; Crop disease; Self-supervised learning; Crop disease; Deep learning\n\n\n# Workshop/conference\n\nThis challenge is part of the conference on \"The International Conference on Pattern Recognition (ICPR) (2024)\".\n\n\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "The evaluation metric for this competition is the categorization accuracy.\n\nThree types are set as follows: Rust, Health and Other\n\n## Submission Format\n\nSubmission files should contain two columns: `Id` and `Category`. \n\nThe file should contain a header and have the following format:\n```\nId\tCategory\n7f4ecf086b6b4c04ae36add24b27063a\tAphid\ne427f07618794fd58dfc9esc786e3743\tRust\n98ceba6c29ab465d9413aa6dfbbeb273\tRPH\n994b5409c8e946538d87209a99897659\tBlast\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "People", "content": "# Organisers\nProf. Liangxiu Han, Manchester Metropolitan University\nProf. Wenjiang Huang, Aerospace Information Research Institute, Chinese Academy of Sciences\n\n# Technical Chairs\nDr. Xin Zhang, Manchester Metropolitan University\nDr. Yingying, Dong Aerospace Information Research Institute, Chinese Academy of Sciences\nTam Sobieh, Manchester Metropolitan University\n\n# Primary contact\nLiangxiu Han - Email: l.han@mmu.ac.uk\nWenjiang Huang - Email: huangwj@radi.ac.cn", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "\nThe \"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. \n\n# Hyperspectral Data Analysis for Crop Disease Classification\n\nThe 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.\n\n# Exploring Synthetic Hyperspectral Data Generation\n\nThe 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.\n\nThe 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.\n\nParticipants 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.\n\nUltimately, 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.\n\n# Workshop/conference\nThis challenge is part of the conference on \"IEEE International Conference on Image Processing, IEEE ICIP 2025\". \n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }