{ "fetched_at": "2026-10-24T22:35:31Z", "source_url": "https://www.kaggle.com/competitions/beyond-visible-spectrum-ai-for-agriculture-2026p2/data", "pages": [ { "name": "data-description", "content": "# Competition: Beyond Visible Spectrum - AI for Agriculture 2026\n\n## Overview\nParticipants are challenged to develop robust **Self-Supervised Learning (SSL)** methods to extract meaningful representations from satellite imagery, which will then be fine-tuned for **Crop Disease Prediction**.\n\nInnovative approaches that leverage **Multimodal** data are highly encouraged to improve classification accuracy under varying environmental conditions.\n\n## 1. Self-Supervised Training Data (Unlabeled)\nThe core dataset consists of unlabeled Sentinel-2 (S2A) time-series imagery, organized by location and acquisition time. This data is intended for pre-training SSL models (e.g., MoCo, SimCLR, MAE) to learn temporal and spectral features of agricultural land.\n\n**Download:** https://www.dropbox.com/scl/fi/d7cmjdd10qlkp9w895me2/s2a.tar.gz?rlkey=xk1pka3q4nokpmj7dilfys5lw&st=xn8g7k53&dl=0)\n\nIf you encounter issues downloading the full `s2a.tar.gz` file due to network instability or size limits, you can download the dataset in split parts using the links below.\n\nWe have updated the link.\n\nPlease download all the following parts:\n\n- [s2a.tar.gz.partaa](https://www.dropbox.com/scl/fo/0ca5g354y1tn8cw7dwk85/AITOGzyxyqCoKJjhzXbuOH4/s2a.tar.gz.partaa?rlkey=h8yirspxk6r312634qabi412m&st=x90erzjv&dl=0)\n- [s2a.tar.gz.partab](https://www.dropbox.com/scl/fo/0ca5g354y1tn8cw7dwk85/AN5iLpgIOwJ6TWNhqAzuDcw/s2a.tar.gz.partab?rlkey=h8yirspxk6r312634qabi412m&st=v6m8n9w7&dl=0)\n- [s2a.tar.gz.partac](https://www.dropbox.com/scl/fo/0ca5g354y1tn8cw7dwk85/AJutAYqAJ8hQ69aIL4RXZyQ/s2a.tar.gz.partac?rlkey=h8yirspxk6r312634qabi412m&st=ed7eqamq&dl=0)\n- [s2a.tar.gz.partad](https://www.dropbox.com/scl/fo/0ca5g354y1tn8cw7dwk85/AP0sa-5mR8mzBoYBN10sa9w/s2a.tar.gz.partad?rlkey=h8yirspxk6r312634qabi412m&st=p0bii2m4&dl=0)\n- [s2a.tar.gz.partae](https://www.dropbox.com/scl/fo/0ca5g354y1tn8cw7dwk85/AG3d3HP7pOgePdY4ni2yJ9s/s2a.tar.gz.partae?rlkey=h8yirspxk6r312634qabi412m&st=odnubqp0&dl=0)\n\n\n### Data Structure\nThe data is organized hierarchically:\nBand_File.tif\n\n* **Band Files**: Individual GeoTIFF files for each spectral band.\n\n### Available Bands (Sentinel-2)\nEach acquisition folder contains the following 12 spectral bands:\n* **Coastal/Aerosol**: `B1.tif`\n* **Visible**: `B2.tif` (Blue), `B3.tif` (Green), `B4.tif` (Red)\n* **Red Edge**: `B5.tif`, `B6.tif`, `B7.tif`\n* **NIR (Near Infrared)**: `B8.tif`, `B8A.tif`\n* **Water Vapour**: `B9.tif`\n* **SWIR (Short-Wave Infrared)**: `B11.tif`, `B12.tif`\n\n\n## 2. Supervised Training Data (Labeled)\nThis dataset contains labeled samples for the downstream classification task.\n\n##Submission Columns\n\n* `Id` - file name\n* `Category` - Aphid, Rust, RPH, Blast\n", "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 **31 Dec, 2025**.\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 **March 1st, 2026**, 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": " **Task 2: Boosting Automatic Crop Diseases Classification using Sentinel Satellite Data and Self-Supervised Learning**\n * Participants are challenged to employ self-supervised learning (SSL) techniques to identify various types of crop diseases from remote sensing imagery.\n * Addresses the issue of limited labelled data in remote agricultural regions.\n * Evaluation will be based on transfer performance, measuring how well the learned representations generalize to new tasks after fine-tuning on a labelled dataset.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Prizes", "content": "Total Prizes Available: 3\n- **1st Place** - £100\n- **2nd Place** - £60\n- **3rd Place** - £ 40\n\n## Organizers\n* **Prof. Liangxiu Han**, Manchester Metropolitan University, United Kingdom\n* **Prof. Wenjiang Huang**, Aerospace Information Research Institute, Chinese Academy of Sciences\n\n**Technical Committee**:\n* Dr. Xin Zhang, Manchester Metropolitan University, UK\n* Dr. Yue Shi, Manchester Metropolitan University, UK\n* Dr. Yingying Dong, Aerospace Information Research Institute, Chinese Academy of Sciences\n\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "# ICPR 2026 Competition on “Beyond Visible Spectrum: AI for Agriculture”\n\n\nThank you all for your incredible efforts and contributions to the Beyond Visible Spectrum: AI for Agriculture 2026 competition. We were truly impressed by the innovative approaches and hard work demonstrated by the community throughout this challenge!\n\nNow that the private leaderboard has been finalized, we would like to officially congratulate our Top winners:\n\nAction Required for Top 10 Teams: To proceed with the reward distribution, we request that the top teams send a detailed write-up of your final solution and methodology to our official email: robeson1010@gmail.com.\n\n@devconsultant , @hoanganhjuu , @miaanguyenn7 , @omowonuola @ethelogallo , @rabinatwayana , @rabina29 , @kushp3690 , @twayana123 , @shallu , @guptaprashant1986 , @tuanhoang777 @duytran2311 @sabaakramofficial , @ewilipsic ,@mrhuang999 ,@savanjaviya , @gauravparkhedkar \n\nDeadline: Please submit your solutions within 20 days (by March 20, 2026).\n\nPlease ensure your email includes the following:\n\nMethodology: A detailed write-up explaining your approach and final solution.\n\nCode & Weights: Your source code and the final model weights (providing a link to a GitHub repository or Google Drive is perfectly fine).\n\nPersonal Details: Your full real name and preferred contact information. (Note: This information is required for creating your certificates and for listing you as a co-author on the official Competition Report/Paper).\n\nNext Steps: Our team will carefully review your methodologies to verify the results and begin drafting the official Competition Reports.\n\nThank you once again to everyone who participated and helped make this competition a massive success. We look forward to reading your solutions!\n\n\n## Abstract\n\nThe ICPR 2026 Competition on “Beyond Visible Spectrum: AI for Agriculture” offers a unique opportunity for researchers to advance computer vision techniques in agricultural crop disease monitoring. Building on a successful series of competitions from 2022 to 2025, which attracted over 1000 participants from over 10 countries across Europe, Asia, and North America.\n\nThe challenge focuses on developing innovative deep learning algorithms using extensive multi/hyperspectral and satellite remote sensing datasets across two main tasks:\n\n* **Task 1: Automated Multimodal Crop Disease Diagnosis from Multimodal Remote Sensing Imagery**: Participants will be provided with RGB, multispectral (G/R/RE/NIR), and hyperspectral imagery. They will be encouraged to develop multimodal models capable of adapting to remote sensing data of varying spectral and spatial information to classify the data into three categories: 'rust,' 'healthy,' and 'other.'\n* **Task 2: Boosting Automatic Crop Diseases Classification using Sentinel Satellite Data and Self-Supervised Learning (SSL)**: Addressing the critical issue of limited labelled data in agriculture, this task requires participants to employ self-supervised learning (SSL) to learn meaningful representations from large, unlabelled Sentinel satellite datasets before fine-tuning on limited labelled data.\n\nThe primary goal is to enhance the accuracy and efficiency of identifying crop diseases, contributing to significant advancements in precision farming and sustainable agricultural practices to support global food security.\n\n\n### Background and Task\nRemote sensing technology offers continuous global monitoring of vast agricultural areas in real-time, enabling early identification and mitigation of threats. Advances in digital imaging have led to various image-based methods for automated plant management, showing great potential for crop disease diagnosis using RGB, thermal, multi-spectral, and hyperspectral sensors. Hyperspectral imagery (HSI), with its numerous narrow spectral bands, provides detailed spectral–spatial information, offering enhanced diagnostic accuracy for disease infestation. However, the high data dimensionality of HSI can also present challenges, leading to research into new methods like super-resolution and automated detection models.\n\nDeep learning methods have shown promising results in remote sensing imagery analysis. Identifying relevant spectral features and extracting meaningful representations from remote sensing data are crucial for effective deep learning. However, the high number of spectral bands and complex relationships in remote sensing images pose significant challenges for feature selection and extraction, necessitating the development of new deep learning algorithms.\n\nBuilding on our research in precision agriculture and extensive experience, we are hosting The ICPR 2026 Competition on “Beyond Visible Spectrum: AI for Agriculture.” This is our fifth consecutive year running this competition. We have seen a remarkable and sustained increase in community engagement, demonstrating the critical importance of this research area.\n\n\n## Competition Schedule\n* **Launch Date**: December 15, 2025\n* **Submission Opening**: December 30, 2025\n* **Deadline for Competition Participants**: March 1st, 2026\n* **Winner Announcement**: March 15, 2026\n* **Initial Submission of Competition Reports Deadline**: April 1st, 2026\n* **Presentation at ICPR Conference**: August 17–21, 2026\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }