Observations validation

Observation validation is a fundamental pillar of Pl@ntNet, ensuring the quality of shared data and its relevance for scientific research. This process combines human contributions and automated calculations to assess the reliability of each observation.

Votes

1. User votes Pl@ntNet users actively participate by voting on observations. Their contribution is weighted by their “weight”, which reflects their experience and reliability within the application.

2. Partner votes Partners, such as institutions or associated experts, have votes with a fixed weight.

3. Calculation of Scores and Probabilities

• Value score: This is the sum of the weights of users who voted for a species.

• Value probability: This is calculated by dividing the score of that value by the sum of the scores of all values proposed for the observation.

Observation Validation Criteria

To be validated, an observation must satisfy several conditions:

1. Valid image

   • The image must not be marked as “not a plant.”
   • It must obtain a quality score of at least 2.

2. Non-malformed observation

   • It must not contain images of several distinct species (“malformed”).
   • The “malformed” vote score must be less than 2.

3. Correct determination

   • The determination score must be at least 2.
   • The probability of the determination must be greater than or equal to 0.7.

4. Absence of censorship

   • The observation must not be censored by the identification engine.

5. Other criteria

   • The user who created the observation must not be blocked.
   • All mandatory fields must be completed (for example, for partner observations).
   • Images must include identifiable organs (such as leaves, flowers, or fruits).

Special Cases

An observation can be validated even if the identified species is not yet indexed in Pl@ntNet’s taxonomic databases. These observations will appear in contribution feeds and user profiles but will not appear in the image galleries. If the species is added later, the observation will be automatically linked to the corresponding gallery.

Importance of Image Quality

To maximize the chances of validation, it is essential to provide high-quality images:

• Take several photos of distinctive organs (leaves, flowers, fruits, etc.).

• Ensure images are sharp and well-framed.

• Enable geolocation to enrich the observation with precise contextual data.

Impact of Validated Observations

Validated observations feed into open databases like GBIF, contributing to research on biodiversity, ecology, and computer science. They also participate in the continuous improvement of Pl@ntNet’s artificial intelligence models, thereby increasing the accuracy of future identifications.

This rigorous process ensures that the collected data is of high quality, reliable, and usable for research, while encouraging collaboration among users.