About the project

Why we do it

Eden Library is a collection of high value plant datasets embedding agricultural domain knowledge produced in an academic environment. Our goal is to use AI capacity in agriculture for food safety and quality control, while adding value to the dataset users. Based on our experience and current trends, we have realised that high quality data is a prerequisite for effective AI tools and for that reason we invest in domain experts and solid processes. We focus on exhaustive annotations and integrate quality control frameworks in order to produce i) problem-specific datasets ii) rich in metadata, iii) clustered based on multiple criteria and iv) standardized in ways that facilitate researchers, tech industry and agrifood business experts to use them in various contexts based on their needs. We aspire to make agriculture cool again, attract young generation and build a strong community in AI agrifood business and research.

Who we are

Started as an initiative from Agricultural University of Athens, Eden Library is a multidisciplinary team of AI experts, agronomists, software engineers committed to close the gap between computer vision and real world agri-food challenges. The team has a vast experience in numerous projects in all aspects of smart farming, robotics in agriculture, artificial intelligence, engineering and multiple collaborations with research centers, universities and multinational companies.

How we collect images

All images are collected under real-life (field & greenhouse) conditions by the Eden Library team and its partners, always making sure that there is a clear distinction of the symptom/plant and that background information is maintained. A variety of sensors are used and lighting conditions are variable, while the collection style is suitable for typical precision farming and robotic tasks. Illumination conditions are intentionally not optimal and not all images have the same resolution and colour balance. Our belief is that such data is needed if AI is going to be widely adopted and used for real life applications.

  • Images are acquired in open field, as well as, greenhouses using a variety of imaging sensors.
  • In-field validation of symptoms by experts.
  • The distance between the plant and the camera is such that allows the clear distinction of the symptom.
  • The number of each dataset is appropriate for prototyping and AI performance enhancement.

How we process and annotate images

All images after being collected are checked and categorized by expert agronomists (e.g. entomologists and phytopathologists) based not only on what is visible in the image but also on additional information from the imaging site. Such information can be the presence of an insect or a sample that was taken to be further studied at our premises. Diseases are identified using additional agronomic techniques, such as cutting the bark. Specific species or diseases are identified, supported by field observations or evidence in the surrounding area.

Images that do not match the experts’ high standards (e.g. symptom is not clear) are discarded. The remaining images are annotated by our in-house trained skilled annotators. Image annotation is the process of adding metadata for the description of an object of interest (e.g. crops, weeds, vegetables) which are then used to “train” machine learning models to detect, recognize, and interpret objects of interest. The quality of the annotations is checked using our own proprietary quality assurance tool. Images that fail this quality control test are sent back to be annotated again. All images and annotations are checked by domain experts before they are uploaded to our database.

Tasks: Specific use cases that match in our point of view with specific datasets e.g. bounding box annotations around symptoms for disease detection, bounding boxes around fruits for harvest prediction, etc. These are not by any means the only use cases and of you are eligible to use the dataset and annotations for your specific scenario.

What kind of datasets we have

Eden Library includes a wide range of agrifood datasets such as:

  • Plant pests
  • Plant diseases
  • Weeds
  • Healthy plants

That were acquired using:

  • Various styles (Proximal, UAV upon request)
  • Various sensors (RGB, thermal, multispectral & hyperspectral upon request)

Can I contribute with my own dataset?

At the moment, contribution from community members or third parties is not possible, because symptoms are examined at field-level by our experts and images are collected in a consistent way to maintain Eden Library’s quality standards. In future versions, we are considering options to allow dataset contribution from community members.

How to download datasets

  1. The process is simple; spend beyond 50 credits you will receive an email reminding you to contact us about the projects you are currently working on and how our datasets are of use to you. Once we talk to you or receive your email explaining how our datasets are helping you reach your goals, we will grant you additional credits.
  2. Every premium dataset comes with a specific price, considering the added value it delivers to the user. For each premium dataset, a “request price” button is available and our team will contact you with more information, upon receiving your request.

How to use the datasets

  1. Please make sure you have read the licensing section carefully in the terms of use, so you do not violate any legal condition
  2. If you use the Eden Library data for your project, please attribute using one of the following options:
    • © Eden Library
    • Citation of the Eden Library paper. When using one of the datasets hosted in Eden Library, please cite our paper. Here is the BibTeX citation entry:
title={Eden Library: A long-term database for storing agricultural multi-sensor datasets from UAV and proximal platforms},
author={Mylonas, Nikos and Malounas, Ioannis and Mouseti, Sofia and Vali, Eleanna and Espejo-Garcia, Borja and Fountas, Spyros},
journal={Smart Agricultural Technology},

Cite the explicit dataset, using the formal name as seen in the details view:

An example of a dataset’s detailed view. You can copy the dataset’s formal name, as seen in the bottom, in order to cite it.


What do they show?

The notebooks available on the Eden library show our modest contribution to the deep learning community interested in precision agriculture problems. They synthesize some of the most common problems and solutions we have addressed in the past years. It is important to remark that the notebooks have not been finely tuned for showing the best possible results; they just present ways to use the datasets in a basic manner. If you want to know more about notebooks as a development tool, please, check this link: What is the Jupyter Notebook?

The notebooks published in the Eden platform are open for extension with additional information, bibliography, and new pieces of source code which could improve the readability of the explained techniques. If you have any suggestions for improving them or any doubts to be clarified, do not hesitate to open an issue in the Github repo. We will reply to you as soon as possible.

Last but not least, we will be very glad if you clone the repository and perform your own experiments and extensions inside the notebooks: Eden-Library-AI_notebooks

In case you can’t visualize the notebooks on Github, check the Nbviewer option, which is also available for a faster visualization:

How to run them?

All the information for running the notebooks is available on our wiki:

There, you are going to find how to run them locally, on Google Colab, by using Docker, and other options. Moreover, the wiki is updated frequently to cover more approaches and to provide support for any possible problem. Please, don’t hesitate to ask for covering other related topics that make our notebooks more useful.

Which technologies are mainly used?

  1. Keras
  2. Tensorflow
  3. OpenCV
  4. Scikit-learn
  5. PyTorch
  6. AutoKeras
  7. Scikit-image

Will new notebooks be created?

Definitely yes! We will update the notebooks section with new notebooks implementing new (and traditional) techniques related to deep learning and image processing which can be directly applied into datasets extracted from the Eden library (and similar ones).

Your personal account in Eden Library

In order to download a free dataset or request for a premium one, you will have to create an account using your email and a password. Some more information regarding your profile are also required in order to help us provide you with the best functionalities based on your needs. All the required information is protected by the privacy policy and terms of use.

Sign-up or download problems

  • “I haven’t received the confirmation email”Please, check the spam folder of the email account you used for signing-up. If you haven’t received it after 10 minutes, please try again.
  • “I have received the confirmation, but I can’t see the link to confirm”Sometimes, some email clients hide or disable links for security reasons. It depends on every email client, but usually, setting the email as a non-junk email should be enough to see the confirmation link.

If the problems persists, please send us an email to We’ll reply to you as soon as possible 😉

  1. “My download is not starting” In many cases, browsers may block by deafult the download confirmation pop-up window which shows up once a user clicks to download one or more datasets. Make sure to allow pop-up windows in your browser, in order to finish the downloading process.

Any more questions?

If you face issues of any kind using Eden Library, please contact us: