Wildlife Insights Artificial Intelligence Models

The Wildlife Insights platform streamlines conservation monitoring by speeding up camera trap data analysis. Researchers who have camera trap data can now upload them to the Google Cloud-based platform, where they can manage, identify and analyze their biodiversity data. Researchers can also run AI models over their data for species identification. Using Google’s open source TensorFlow framework, Google has trained AI models to perform two tasks. 

1. Blank image filtering: We heard from wildlife researchers that many hours are spent doing the arduous manual task of eliminating blanks from a dataset;  that is, filtering out images without animals. Our models are tuned to accurately classify blanks, while reducing the possibility of removing valuable images of animals. We set a very high threshold on the required confidence of the model when predicting the blank class, so if we do classify it as a blank, the probability that it is actually a blank is very high. 


2. Species Classification: We trained our AI models to learn to recognize hundreds of species from around the world. Wildlife Insights is an unprecedented partnership among organizations who have spent decades collecting and cataloguing or labelling animals in camera trap images. This is a high-quality training dataset for AI models, consisting of over 8.7M images, larger than all the world’s vertebrate specimen in museums combined. 

We need your help! Our current dataset is the largest curated dataset of labeled camera trap images to date, with examples of 614 species. There are many more vertebrate species in the world. You can help us expand our AI capabilities to new species or existing rare species, and improve our accuracy on the species already in the dataset, by adding your data to Wildlife Insights. You can benefit from others sharing data as the models get better and better for the species you care about.

We are driven by a mission to support people who care about preserving wildlife for current and future generations. Our goal is to provide insights for effective conservation management decision-making. We hope you enjoy Wildlife Insights!
 

FAQs about Wildlife Insights AI

What data was used to train the Wildlife Insights AI models, and which species can the model identify?

Aside from identifying species, how else can Wildlife Insights AI models help optimize human work reviewing camera trap images?

How accurate are the Wildlife Insights AI models at identifying animals in images?

Will the Wildlife Insights AI models correctly predict the species I care about?

How does the model decide on a particular class?

What parts of the image are important when predicting a class?

What happens when new images are uploaded to Wildlife Insights?

How fast will you return results if I upload 10 or 10,000 images?

What does the “confidence” mean next to an image’s classification?

Will Wildlife Insights AI models be updated?

What taxonomy does Wildlife Insights use?

Can AI identify individuals from species with spots or stripes "naturally marked" animals: e.g., zebras or tigers?

Will Wildlife Insights support videos taken by camera traps?

How can I get Wildlife Insights to analyze my camera trap data?

Does Wildlife Insights analyze humans as a class in the AI models?

Can I help improve the AI models with my camera trap data?

What other plans do you have to improve the AI models?

What data was used to train the Wildlife Insights AI models, and which species can the model identify?

The AI model is trained on images from Conservation International's Tropical Ecology and Monitoring (TEAM) Network, Snapshot Serengeti, Caltech Camera Traps, North American Camera Trap Images, WWF and One Tam, which include 614 species from around the world.

We are adding to this core training dataset frequently on an ongoing basis with data from Wildlife Insights core members: Conservation International, Smithsonian Institution, North Carolina Museum of Natural Sciences, Wildlife Conservation Society, ZSL, and WWF. We have also trained using openly available datasets on lila.science.

Our training data contains images labeled with WI taxonomy including Class > Order > Family > Genus > Species. Species, which is the most granular level, is used as a class label to train a multi-class image classifier.

We are using a deep convolutional neural net for multi-class classification using Google’s open source TensorFlow framework to train an AI model to identify animal species in camera trap images. 

Like humans, AI models generally get better at recognizing and identifying animals if they can look at hundreds or thousands of diverse images of that particular species. If you have camera trap data with many images of popular species or even a few images of rare species, we encourage you to contact us to get trusted tester access to Wildlife Insights, so that you can more easily manage and identify your camera trap images and contribute to the accuracy of Wildlife Insights AI models. 

Aside from identifying species, how else can Wildlife Insights AI models help optimize human work reviewing camera trap images?

Perhaps most importantly, AI can help identify images without animals where things like blowing grass can trigger a camera, which can be up to 80% of a dataset. Leveraging Google AI Platform Predictions, this functionality alone can dramatically reduce the amount of time spent processing and identifying camera trap data. 

The first task for the AI models has been to identify images not containing any animals, since no one wants to look at thousands of empty images. The AI models in Wildlife Insights catch 78.7% of blank images with an error rate of less than 2%. This allows Wildlife Insights users to minimize the amount of human involvement in the process of sifting through millions of images looking for wildlife, and allowing tens to hundreds of machines to do that work for them in a fraction of the time. Upload your images, put computer vision to work for you, and when the results come back you can focus on the images that need your attention. 

How accurate are the Wildlife Insights AI models at identifying animals in images?

Across the 614 species that the models have been trained on classes like blue duiker, African elephant, southern pig-tailed macaque or a small antelope called Suni have between 80% and 98.6% probability of being correctly predicted by the AI models. 

Categorizing species in camera trap data can be very challenging, even for humans.  Data quality can play a huge part in our ability to correctly classify an image, and even human experts struggle with images that are poorly illuminated, blurry, or where the animal is very small, hidden behind vegetation, or far away.  There are also many sets of species that are easily confused, like bobcats and lynx. Images that have low data quality or contain easily-confused species are harder for both humans and AI.  That said, AI improves when given many, diverse examples of a given species.  

This is where you can help!  You can correct any mis-labeled species using the Wildlife Insights interface and improve the model accuracy for that species class. In addition, by adding your data to Wildlife Insights, you can help provide sufficient examples of each species to our AI systems so that the species can be accurately identified in the future. By uploading data from your unique camera traps you are not only improving species accuracy.  Each individual camera trap has a set of biases, such as the background, perspective on the animals, and lighting conditions. Your data is also increasing the camera diversity of our dataset, and in turn improving the robustness of our AI models to varied camera conditions.

Will the Wildlife Insights AI models correctly predict the species I care about?

Once you’ve uploaded your camera trap images to Wildlife Insights, you will see the per-image classification confidence displayed alongside the image in the Identify section of Wildlife Insights. 

You can also look up individual species to see if that species has examples in our current training dataset, and what the model performance is on that class. 

Search for a species of your interest. Let’s take an example of mule deer. If there were total 100 images of mule deer in the data you uploaded, we will be able to identify 81 of those as mule deer. This is the recall for the class. If we predict 100 images as mule deer, about 93 of them are likely to be actually mule deer and the other 7 are likely something else. That is the precision for the class.

If you see “Needs More Data” in the metrics, this means our AI model is not able to predict a single class with confidence above a fixed threshold (threshold is tuned manually). This may happen due to a low number of images of this species in training, due to less diversity in the images (e.g. all from same camera location, similar background etc.), or just because the characteristics to identify the species is shared between many species which confuses the model. You can help improve our model accuracy by contributing additional data to Wildlife Insights for your species of interest. As users upload more images - from different regions, more diverse species etc., our AI models will get better at recognizing more species.

If you do not see your species of interest listed it means that we currently have no examples of that species in our dataset. This is all the more reason to contribute, so we can continue to grow the number of supported species in Wildlife Insights.

Learn more about assessing classification accuracy for AI models in general.