iForum LIVE!: Why are we integrating FASTCAT-Cloud and the Plantnet API into iSpot?

These 13 services will be upload to the EOSC , so that any existing citizen science observatory will be able to choose and install the technological services needed to improve its functionalities**.**

2 Likes

You can find infographics and detailed information about the services on the following website: Cos4Cloud Services – Cos4cloud

For example, for Pl@ntNet API you could explore some initial information :point_right: Pl@ntNet-API – Cos4cloud

3 Likes

In this context EOSC stands for European Open Science Cloud, a virtual space for the European scientific community…

1 Like

Thanks @adrisoacha . Please do post any questions you have for Karen. Next @srueger will tell us about FASTCAT-Cloud.

FASTCAT (Flexible Ai System for CAmera Traps)-Cloud is an open website service https://service.fastcat-cloud.org/ where videos or images of wildlife observations can be uploaded for the purpose of species identification and/or filtering out empty images automatically. The service uses Artificial Intelligence trained on wildlife observations from GBIF for a particular region or theme. Currently FASTCAT-Cloud has a model for 36 land-based mammals in the UK and one for 22 mammals in California, but other models are currently being trained, eg, one with 268 UK birds.

Uploading can be done interactively by a user through a website or by automated processes/computers through an API (application programmer interface), for example, by internet-enabled camera traps, by biodiversity monitoring stations, or by nature enthusiast networks like iSpot.

The API service allows others to process their imagery automatically, eg, to filter out empty images from a camera trap, make species ID predictions or count number of sightings over a longer period etc

1 Like

We would love to hear from you with any questions you have for us iSpotters…

Does the system allow other types of camera traps to be connected and filter images and identify for those too?

2 Likes

Absolutely, as long as the camera trap has internet and a means to push to the API - it;s all an open system for use by anyone

2 Likes

Unfortunately Alexis from Pl@ntNet is a delayed in joining us so will pick up on the discussion later. In the interim I will tell you a bit about the Pl@ntNet-API.

2 Likes

Sounds like a great Raspberry Pi project!

3 Likes

In fact, the camera traps that we use are based on the RPi. Have a look at FASTCAT-Edge

2 Likes

Pl@ntNet-API enables other citizen observatories and third-party applications developed by industrial, academic or associative stakeholders to integrate automatic plant identification features very easily into their apps. identification engine. This service is available to developers, researchers and other citizen science observatories interested in plant biodiversity

Pl@ntNet-API enables other citizen observatories and third-party applications developed by industrial, academic or associative stakeholders to integrate automatic plant identification features very easily into their apps. This service is regularly updated and enriched. Because it is connected to the Pl@ntNet database, it is regularly updated with new flora images. This means the visual recognition is constantly improving.

Additionally, Pl@ntNet-API incorporates automated rejection of inappropriate content. This means that any pictures of non-plant entities are automatically tagged REJECT. Specifically, inappropriate content such as faces or pornographic content receives special attention for recognition.

The service also monitors the quality of service provided based on several criteria (status, response time, scalability, etc.) and includes an alert system that allows developers to react quickly.

How it works: A citizen observatory developing a new app related to biodiversity will be able to integrate an automatic plant identification feature without managing it itself.

**Guidelines and documentation to start using Pl@ntNet-API:**The website my.plantnet.org provides rich documentation of the service as well as a dashboard allowing users to track their use of the service.

2 Likes

@mathias.chouet do you have anything further to add?

Do you currently use AI in any capacity at all to help with recording and identifying?

@miked why are we integrating these technologies into iSpot?

That is a good question Janice and one that you might want to add some more info about.

iSpot is experimenting with automatic image identification to help the community identify observations and to help with learning. The main method of identification is still by the community giving identifications and helpful comments.

One of the main reasons I use systems such as this is when I know the species but have forgotten the name, my memory has never been good for names even when I know what it is.

On other occasions it may be a species that I am not familiar with at all, for example from a different part of the world. On these occasions you have to be very careful with the AI and remember that it only knows about species that it has been trained on so you have to check that it looks right but also that there are not other similar species which may not be included in the AI system. This can be particularly the case with garden plants where cultivars can look quite different to the wild species.

When you come across a species that you are not familiar with at all it can be like being a beginner again and think about asking what would be helpful for a beginner. Would you want to immediately give them just the answer or is it better to give a range of possibilities. If it were a human teaching them then they may suggest characteristics to look out for and point out similar species in the genus or family. Could the AI do some of this, perhaps with natural language in addition to just the suggestion in the comments. iSpot does have the species browser and taxonomic tree and huge numbers of correctly identified images so could potentially give some of this additional information automatically e.g. via a link.

4 Likes

Here is an example where the Plantnet AI gets the ID wrong but where a human could easily do the ID Water Mint | Observation | UK and Ireland | iSpot (ispotnature.org) There are interesting reasons why the AI fails here even though it works in the vast majority of cases, for example it does not take into account the habitat or other informatoin that the user has supplied

2 Likes

I know there is a lot of information coming in here but if anyone has any questions, please just drop them in! :slight_smile:

And that’s a particularly good plant observation because it has more than one image for the AI to chew on.

1 Like

Good point @Chris_Valentine . Can you tell us how are we integrating these technologies into iSpot?