\nThe training data is composed of the species observations (output) and environmental data (input). We explain below their precise structure and data sources.\n
\n
The training species observation data is composed on one hand of more than 5 millions of plant species presence-only occurrences (species name, geolocation, time, dataset, observer, etc. ) collected since 2017 in Europe (38 countries), and on the other hand of around 5 thousands presence-absence surveys (inventory of all species in a small plot of 10 to 400m²). This data covers overall around 10,000 species, namely most of the European flora. This data should be used to inform the model outputs keeping in mind that the presence-only data (PO) only partially represent the local species composition and it is subject to various sampling biases. The local absence of a species among PO data doesn't mean it is truly absent: An observer might not have reported it because it was difficult to detect at this time of the year, difficult to identify, not a monitoring target or just not attractive. Indeed, most of the PO data come from citizen science programms without sampling protocol. The presence-absence data (PA) is provided to compensate the problem of false-absences of PO data and calibrate models in order to avoid associated biases.
\n\nSpecies observation details. In more details, there are two occurrence CSV for the training species observation data, one containing the PO data (/Presence_only_occurrences/Presences_only_train.csv) and the other the PA data (/Presence_Absence_surveys/Presences_Absences_train.csv). Both CSV include notably the columns speciesID (species are anonymised), lon (longitude WGS84), lat (latitude), geoUncertaintyInM, year, dayOfYear (1 to 366), glcID (observation id), gbifID (for gbif data only, matches with www.gbif.org), patchID (see below), timeSerieID (see below), datasetName (string) and observer (string). Hence, with PA data, there is generally many speciesID for a same combination of geolocation, year and dayOfYear, and potentially several observers or datasetNames.
\n\nWe also provide spatialized geographic and environmental data to be used as input variables. For each species observation, we provide a four-band 128x128 satellite image at 10m resolution around the occurrence location and time series of the past values for six satellite bands at the point location. Besides we provide various environmental rasters at the European scale (e.g., climatic and pedologic, land use and human footprint variables) from which you may extract local 2D images and build 3D tensors using the dataloader provided on our Github. These three types of input environmental data are illustrated in the Figure below. Note that you may also extract time series of climatic variables for any observation, since we provide monthly climatic rasters from 2000 to 2019.
\n\n\n![](\"https://filedn.com/ltHjVlFrSfSJL1S4hXqqTzy/Figure.jpg\")
Figure. Illustration of of the environmental data for an occurrence (glcID=4859165) collected in northern Switzerland (lon=8.5744;lat=47.7704) in 2021. A. The 1280x1280m satellite image patches sampled in 2021 around the observation. B. Quarterly time series of six satellite bands at the point location since winter 1999-2000. C. Three example bioclimatic images (~65x65km) around the observation, extracted from the provided environmental rasters.\n
In more details:\n
We noticed that the download of data folders in seaFile is long when they contain large files. This is due to a compression of the folders prior to download. We will soon share a few archives containing the whole dataset for direct download. Note that you can still directly download the files without compression by using the download button on the right of each file row.
\n\nTo download the full dataset, we recommend to download the following (groups of) folders independently and according to your needs (pre-compression time on seaFile in brackets):\n
Then, directly download the following files (to avoid the pre-zipping) by clicking on the direct download button on the right of the file name in the seaFile repository:\n
\nContinuously predicting the composition of plant species and its change in space and time at a fine resolution is useful for many scenarios related to biodiversity management and conservation, the improvement of species identification and inventory tools, as well as for educational purposes.
\n
\nThe objective of this challenge is to predict the set of plant species present in a given location and time using various possible predictors: satellite images and time-series, climatic time-series, and other rasterized environmental data: land cover, human footprint, bioclimatic and soil variables.
\n
\nTo do so, we provide a large-scale training set of about 5M plant occurrences in Europe (single-label, presence-only data) as well as a validation set of about 5K plots, and a test set with 20K plots, with all the present species (multi-label, presence-absence data).
\n
\nThe difficulties of the challenge include: multi-label learning from single positive labels, strong class imbalance, multi-modal learning, large-scale.\n
![\"graphical](\"https://filedn.com/ltHjVlFrSfSJL1S4hXqqTzy/GLC_2023_graphical_abstract.jpg\")
\n\n\n Predicting the set of plant species present at a given location is useful for many scenarios related to biodiversity management and conservation.
\n
\nFirst, it allows building high-resolution maps of species composition and of related biodiversity indicators such as species diversity, presence of endangered species, presence of invasive species, etc. In scientific ecology, the problem is known as Species Distribution Modelling.
\n
\nMoreover, it could allow to significantly improve the accuracy of species identification tools - such as Pl@ntNet - by reducing the list of candidate species observable at a given site.
\n
\n More generally, it could facilitate biodiversity inventories through the development of location-based recommendation services (e.g. on mobile phones), encourage the involvement of citizen scientist observers, and accelerate the annotation and validation of species observations to produce large, high-quality data sets.
\n
\n Finally, this could be used for educational purposes through biodiversity exploration applications with features such as quests or contextualized educational pathways.\n
\n This competition is held jointly as part of:\n
\n Being part of scientific research, the participants are encouraged to participate to both events.
\n In particular, only participants who submitted a working note paper to LifeCLEF (see below) will be part of the officially published ranking used for scientific communication.\n
\n LifeCLEF lab is part of the Conference and Labs of the Evaluation Forum (CLEF).
\n CLEF consists of independent peer-reviewed workshops on a broad range of challenges in the fields of multilingual and multimodal information access evaluation, and a set of benchmarking activities carried in various labs designed to test different aspects of mono and cross-language Information retrieval systems.
\n CLEF 2023 will take place in Thessaloniki, Greece, 18-21 September 2023.
\n More details can be found on the CLEF 2023 website.\n
\nParticipants should register the LifeCLEF 2023 lab using this form (and checking \"Task 3 - GeoLifeCLEF\" of \"LifeCLEF\" section).
\n This registration is free of charge.\n
\nThis registration will give you the opportunity to present your results to the CLEF community, during the LifeCLEF session of CLEF 2023, if you win the challenge and submit a working note.
\nIndeed, participants are required to submit, at the end of the competition, a working note paper to LifeCLEF which will be peer-reviewed and published in CEUR-WS proceedings.
\n This paper should provide sufficient information to reproduce the final submitted runs.
\n
\n Submitting a working note with the full description of the methods used in each run is mandatory.
\n Any run that could not be reproduced thanks to its description in the working notes might be removed from the official publication of the results.
\n Working notes are published within CEUR-WS proceedings, resulting in an assignment of an individual DOI (URN) and an indexing by many bibliography systems including DBLP.
\n According to the CEUR-WS policies, a light review of the working notes will be conducted by LifeCLEF organizing committee to ensure quality.
\n As an illustration, LifeCLEF 2022 working notes (task overviews and participant working notes) can be found within CLEF 2022 CEUR-WS proceedings. \n
\n This competition is part of the Fine-Grained Visual Categorization FGVC10 workshop on the 18th June at the Computer Vision and Pattern Recognition Conference CVPR 2023.
\n A panel will review the top submissions for the competition based on the description of the methods provided.
\nThe results of the task will be presented at the workshop and the contribution of winner team(s) will be highlighted. Attending the workshop is not required to participate in the competition.\n
\n CVPR 2023 will take place in Vancouver, CANADA, 18-22 June 2023.
\n PLEASE NOTE: CVPR frequently sells out early, we cannot guarantee CVPR registration after the competition's end.
\n If you are interested in attending, please plan ahead.\n
\n You can see a list of all of the FGVC10 competitions here.\n
\n\nQuestions can be asked in the discussion or to geolifeclef@inria.fr
\n\n\n The evaluation metric for this competition is the micro \\(F_1\\)-score computed on the test set made of species presence-absence (PA) samples. In terms of machine learning, it is a multi-label classification task. The \\(F_1\\)-score is an average measure of overlap between the predicted and actual set of species present at a given location and time.
\n Each test PA sample \\( i \\) is associated with a set of ground-truth labels \\( Y_i \\), namely the set of plant species (=speciesId) associated with a given combination of the columns patchID and dayOfYear (see the Data tab for details on the species observation data structure).
\n For each sample, the submission will provide a list of labels, i.e. the set of species predicted present \\( \\hat{Y}_{i,1}, \\hat{Y}_{i,2}, \\dots, {\\hat{Y}}_{i,R_i} \\).
\n The micro \\(F_1\\)-score is then computed using\n
\n \\[\n F_1 = \\frac{1}{N} \\sum_{i=1}^N \\frac{\\text{TP}_i}{\\text{TP}_i+(\\text{FP}_i+\\text{FN}_i)/2} \\\\ \\quad \\text{Where} \n\\begin{cases}\n \\text{TP}_i =\\text{ Number of predicted species truly present, i.e. }|\\hat{Y}_i \\cap Y_i |\\\\\n \\text{FP}_i =\\text{ Number of species predicted but absent, i.e. } |\\hat{Y}_i \\setminus Y_i | \\\\\n \\text{FN}_i =\\text{ Number of species not predicted but present, i.e. } | Y_i \\setminus \\hat{Y}_i |\\\\\n \\end{cases}\n \\]\n
\n\n\n![\"Illustration](\"https://filedn.com/ltHjVlFrSfSJL1S4hXqqTzy/split_points.png\")
\n\n\n In order to limit the spatial bias during evaluation, the presence-absence data (PA) were split into validation and test sets using a spatial block holdout procedure, and the presence-only data (PO) were filtered to remove PO near test samples.
\n This procedure is illustrated in the previous figure: The test samples - in blue - are located in randomly drawn blocks of a large spatial grid while the other blocks constitute the validation PA samples - in red.
\nThe train set is fully made of PO samples, i.e. each is the record of one species at a certain location and date, while other species might have been present. Nevertheless, PO samples falling at the exact location of test PA sample would inform on its composition. Hence, we have filtered all PO samples near the test samples, inside a radius of a few hundred meters.\n
\n We provide XX baselines in the leaderboard:\n
\n The submission format is a CSV file containing two columns for each sample (row):\n
Id column containing integers corresponding to the test sample ids, corresponding to unique combinations of patchID and dayOfYear column values. Predicted column containing space-delimited lists of the predicted species identifiers (column spId in training/validation datasets)\n The file should contain a header and have the following format:\n
\nId, Predicted\n1,1 52 10231\n2,78 201 1243 1333 2310 4841\n...\n\n\nFor each sample (row), the predicted species identifiers must be ordered by increasing value from left to right. No test sample is empty and the test set only contain species that are present in the train or validation set.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "## One account per participant\n\nYou cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\n\n\n## No private sharing outside teams\n\nPrivately sharing code or data outside of teams is not permitted. It is okay to share code if made available to all participants on the forums.\n\n\n## Team Mergers\n\nTeam mergers are allowed and can be performed by the team leader. In order to merge, the combined team must have a total submission count less than or equal to the maximum allowed as of the merge date. The maximum allowed is the number of submissions per day multiplied by the number of days the competition has been running.\n\n## Team Limits\n\nMaximum team size is 5.\n\n\n## Submission Limits\n\nYou may submit a maximum of 3 entries per day. You may select up to 5 final submissions for judging.\n\n\n## Competition Timeline\n\nCompetition Timeline dates are reflected on the competition’s Overview > Timeline page.\n\n\n# COMPETITION-SPECIFIC TERMS\n\nCompetition Name (the 'Competition'): GeoLifeCLEF 2023 - [LifeCLEF 2023](https://www.imageclef.org/LifeCLEF2023) x [FGVC10](https://sites.google.com/view/fgvc10)\n\nCompetition Sponsors: Horizon Europe project [MAMBO](https://www.mambo-project.eu/), [FGVC10](https://sites.google.com/view/fgvc9), [LifeCLEF 2023](https://www.imageclef.org/LifeCLEF2023) and Inria\n\nCompetition Acknowledgements: Alexis Joly, Christophe Botella, Benjamin Deneu, Diego Marcos, Théo Larcher, Joachim Estopinan, César Leblanc, Maximilien Servajean (hosts)\n\nCompetition Website: [Kaggle](https://www.kaggle.com/competitions/geolifeclef-2023-lifeclef-2023-x-fgvc10/), [LifeCLEF 2023](https://www.imageclef.org/GeoLifeCLEF2023) and [FGVC10](https://sites.google.com/view/fgvc10/competitions/geolifeclef2023)\n\nWorkshop Sponsors: Horizon Europe project [MAMBO](https://www.mambo-project.eu/), COS4CLOUD and Inria (LifeCLEF 2023), Visipedia and Google (FGVC10)\n\n\n## INDIVIDUALS AND TEAMS.\n\nA. Individual Account.\n\nYou may make Submissions only under one, unique Kaggle.com account. You will be disqualified if you make Submissions through more than one Kaggle account, or attempt to falsify an account to act as your proxy. You may submit up to the maximum number of Submissions per day as specified on the Competition Website.\n\nB. Teams.\n\nIf permitted under the Competition Website guidelines, multiple individuals may collaborate as a team (a \"Team\"); however, you may join or form only one Team. Each Team member must be a single individual with a separate Kaggle account. You must register individually for the Competition before joining a Team. You must confirm your Team membership to make it official by responding to the Team notification message sent to your Kaggle account. Team membership may not exceed the Maximum Team Size stated on the Competition Website.\n\nC. Team Merger.\n\nTeams may request to merge via the Competition Website. Team mergers may be allowed provided that: (i) the combined Team does not exceed the Maximum Team Size; (ii) the number of Submissions made by the merging Teams does not exceed the number of Submissions permissible for one Team at the date of the merger request; (iii) the merger is completed before the earlier of: any merger deadline or the Competition deadline; and (iv) the proposed combined Team otherwise meets all the requirements of these Rules.\n\n## External Data.\n\nDo not cheat. \n\nThe general rule is that participants should only use the provided training and validation data for training models to classify the test images. We do not want participants crawling the web in search of additional data. Pretrained models may be used to construct the algorithms from publicly available academic datasets (e.g., ImageNet, GeoLifeCLEF 2020-2022). Please specify any and all external data and/or models used for training when uploading results.\n\nParticipants are allowed to collect additional annotations on the provided training sets. Participants are not allowed to collect annotations on the test set. Teams should specify that they collected additional annotations when submitting results.\n\n## DETERMINING WINNERS.\n\nEach Submission will be scored and ranked by the evaluation metric stated on the Competition Website. During the Competition Period, the current ranking will be visible on the Competition Website's public leaderboard. The potential winner(s) are determined solely by the leaderboard ranking on the private leaderboard, subject to compliance with these Rules. The public leaderboard will be based on the public test set and the private leaderboard will be based on the private test set.\n\nIn the event of a tie, the Submission that was entered first to the Competition will be the winner. In the event a potential winner is disqualified for any reason, the Submission that received the next highest score rank will be chosen as the potential winner.\n\nParticipants are encouraged to register to LifeCLEF 2023 lab and submit a working note describing your methodology to maximize their visibility. This can be done using this form (and checking \"Task 3 - GeoLifeCLEF\" of \"LifeCLEF\" section). This registration is free of charge. This registration will give you the opportunity to present your results to the CLEF community, during the LifeCLEF session of CLEF 2023, if you win the challenge AND submit a working note. By doing so, you are guaranteed that your results will be documented in the vhallenge overvieww publication.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Resources", "content": "# Resources\n\nBesides this Kaggle page, make sure to check these other resources:\n- [GitHub repository](https://github.com/plantnet/GLC) providing useful data loading. We will soon update it with visualization utilities as well as starter code;\n- [LifeCLEF 2023 webpage](https://www.imageclef.org/LifeCLEF2023) for more information about LifeCLEF challenges and working notes submission procedure;\n- [FGVC10 page](https://sites.google.com/view/fgvc10) for more information about the FGVC10 workshop;\n- Our protocol note explaining the data sources and procedure used to build the dataset will soon be published and made available on this page, stay tuned.\n- Last editions winning solutions working notes:\n + [GeoLifeCLEF 2021 winner, S. Seneviratne](http://ceur-ws.org/Vol-2936/paper-140.pdf);\n + [GeoLifeCLEF 2022 2nd, B. Kellenberg et T. Devis](https://ceur-ws.org/Vol-3180/paper-167.pdf);\n- [Overview paper of GeoLifeCLEF 2022](https://ceur-ws.org/Vol-3180/paper-155.pdf) (/!\\ different dataset and evaluation mode /!\\ ).\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Timeline", "content": "
\n All deadlines are at 11:59 PM UTC of the corresponding day unless otherwise stated.
\n The competition organizers reserve the right to update the contest timeline if they deem it necessary.\n