.DVYMrsEo_9fOI6.webp){kind=logo}
Record frog calls and our web app will tell you the species. Clip it, Ribbit!
Our mission
Empowering conservation through cutting-edge technology, our mission is to bridge the data gap in biodiversity by harnessing machine learning for real-time identification of frog calls.
With a focus on biodiversity hotspots in the global south, we aim to enhance conservation efforts, protect endangered species, and support ecosystem resilience through collaborative, data-driven insights.
What makes Ribbit different?
While citizen science apps have been around for a while, none have focused on amphibians. Amphibians are key to conservation policies, as they are one of the most endangered vertebrate groups in the world, with more than 40% of the species endangered to extinction, partly due to the climate crisis.
We created Ribbit to give nature enthusiasts the possibility to learn more about amphibians, and experts the chance to quickly identify frog species by leveraging AI, in order to make informed decisions for conservation.
Specifically, we wanted to focus on decreasing the biodiversity data gap, encouraging users to record frogs in the Global South. For this reason, our web app can be used in English, Spanish, Portuguese and Arabic.
Targeting the Global South is important, as many biodiversity hotspots (areas rich in nature and threatened) are found in that region.
The tech behind the app
Few-shot transfer learning is a technique in artificial intelligence that allows a pre-trained model to quickly learn and adapt to new tasks using only a small number of examples. It's like teaching a child to identify a cat, by only showing it three images of your pet tabbies. This approach is particularly useful when identifying species for which there is not a lot of data, such as species in the Global South.
Ribbit used the BirdNet model, developed by the Cornell Lab of Ornithology, and based on the ResNet18 CNN model to extract embeddings (a compact representation of the frog's calls) from frog recordings, and then added two classification layers to our embeddings. The first one identified whether the call was coming from a frog or not, and the second classified the call into five possible species. If the model wasn't confident enough to predict a species, it would output the family. For example, if it wasn't sure that the frog was a red-eyed tree frog that was calling, it would output tree frog. We used two datasets to train our model: [iNat sounds](https://github.com/visipedia/inat_sounds) and [Anuraset](https://github.com/soundclim/anuraset).
How to use Ribbit
Follow this walk-through to learn how to use our app, or follow the link to our demo. When you're ready, click the "Ribbit" button to visit our app.
Getting started
Amphibians are one of the most endangered vertebrate groups in the world, with more than 40% of the species endangered to extinction. When using Ribbit, be sure to follow these guidelines to help keep them safe:
- Keep very still while recording.
- Don’t talk while recording.
- Start the recording as soon as the frog starts calling.
- Don’t touch the frogs with your bare hands, to avoid contaminating the species with the chytrid fungus Batracochytrium dendrobatidis, a fungus affecting frogs worldwide.
- After you’re done frogging, please wash your hiking boots or shoes in a solution with water and chlorine, to prevent the spread of the fungus.
Using Ribbit as an iPhone shortcut
You can add Ribbit to your home page by clicking the link below. This will create an iPhone shortcut, so that next time you say "Hey Siri, Ribbit", it will take you straight to our app. Or, you can access it from your home screen, as seen on the image to the right:
Using Ribbit on your phone or computer
1. Choose your language
Ribbit seeks to reduce the biodiversity data gap in the global south, which is why you can select from four languages to do your recordings in. Select your favorite language from the drop-down menu: English, Spanish, Portuguese or Arabic.
2. Create an account
We value your privacy, which is why we want you to be the only one who has access to your recordings. Create a unique Username and Password. You'll receive an email, asking you to verify your id in order to start using the app.
3. Record your frog or upload a recording
- Hearing a frog right now? Click Record Now!
- Do you want to upload a recording? No problem, you can click Upload and Locate Yourself
Learning from frogs
Ribbit not only records frogs, it also lets you learn from them
My Recordings
Keep a record of all the frogs and toads you've recorded in time. You can add notes about them as well, such as where they were calling from, or notes on behavior. Here's some ideas to get you started:
- Was it calling from a tree? Near a pond? Inside your house?
- Were there frogs from other species calling at the same time?
- Were there many frogs of the same species calling at the same time?
My Frogs
Ribbit gives you options when you're recording; it will give you five possible species of frogs which may be calling in your recording. This is because some species of frogs tend to call at the same time, so we want to be able to identify who is calling. In this section you'll find the top 5 species predicted to be calling in each of your calls.
Explore Species
While our current app can identify 71 species of frogs and toads, we have loads more information about other species. Learn about the > 800 frog species in our database. You'll get access to their Wikipedia bio, as well as:
- Common Name
- Family
- Threat Status
- Trade Information
Meet the team
Our team developed Ribbit as our capstone project for the Masters in Information and Data Science (MIDS) program at UC Berkeley.
FAQ
How do I contact the Ribbit team?
You can reach us at julianagc@berkeley.edu
What is Ribbit and how does it work?
Ribbit is a web application for automatic identification of frogs and toads. It works by using AI to identify frog calls and classify them into species.
How is the Ribbit dataset shared?
If you choose to share your recorded calls to biodiversity repositories, then the data will be uploaded to the Global Biodiversity Information Facility and, if the frogs are recorded in Colombia, to Biomodelos.
Terms of Use & Privacy Policy
You can read our terms of service here and privacy policy here
How do I access all my Ribbit records and audio?
You can find all your recordings under the "My Frogs" section in our web application
How does the app tell what frog I have recorded?
We use a combination of a Convolutional Neural Network (BirdNET, created by the Cornell Label of Ornithology) and a linear classification layer to train a model to recognize 44 frog species, by using data collected by users like you, as well as by scientists. If you want to learn more about the data used to train the model, you can visit iNaturalist sounds and Anuraset
Can I submit an image for identification?
No, Ribbit only classifies species by sound. If you were able to capture an image, you can upload it to iNaturalist, which focuses on image-based classification of species.
Can I upload external recordings to Ribbit?
Yes, you can! Just select the button on the landing page and upload your recording, once you have a working connection.
Do I need to have phone reception to make a recording?
No, you can record the frog on your phone, and then upload it to our web app.
Are my recordings public?
You can choose to share your recordings by accepting the Privacy Policy. If you don't want to share them, just unselect the box, and select it again when you want to share the audios again.
When should I record frogs?
Frogs love water, so if you want to take a walk right after a strong shower, you're more likely to hear a frog. Most frogs call at night, but some call during the day, especially in the tropics. You can explore the frogs in our library to find out the perfect time to record a specific frog.
Where are the best places to record frogs?
Since frogs are amphibians, part of their lifecycle is tied to freshwater. So if you're close to a pond, lake, river or stream, you're sure to find them. But these bodies of water can be smaller, places commonly known as "plant ponds"; that is, small water bodies within plants, like in tree trunks or in plants like bromeliads.
I'm not sure if what I'm hearing is a frog, a bird or a cricket? Should I still record it?
Yes! Ribbit can tell them apart. But if you want to know more, frog calls are usually continuous, because they produce the calls by filling air sacs with air, then pushing it out and refilling it, so that they can produce a long call, whereas birds tend to have shorter calls.
Why is it important not to touch frogs and to wash footwear between areas?
Amphibian populations are declining in part due to a chytrid fungus, which humans inadvertently carry from place to place in our footwear. By washing our hiking boots or shoes with a mixture of water and chlorine, we kill off the fungus, so it won't spread to new populations.
Why is the Ribbit app available in only four languages?
While English is a universal language, the Global South has a data gap in terms not only of biodiversity, but also of content generated in these languages. We wanted to decrease this gap, by adding some of the languages most widely spoken in the global south, and less represented. As native speakers of these four languages, we chose English, Spanish, Portuguese and Arabic as a starting point, but hope to add languages further on.