Home

##Animal Wildlife Estimator using Social Media

Welcome to the AnimalWildlifeEstimator wiki!
For any questions reach out to Sreejith Menon(smenon8@uic.edu)

Need for a wildlife population estimator using Social Media images:

Manual estimation of animals in wild is troublesome and is known to be prone to errors. The guidelines for deciding if a particular species is endangered or critically endangered within a geographic region is often very stringent. (Reference: Categories and Criteria.)
Thus, there is an urgent need for accurate predictions of population of species in different geographical regions.

IBEIS:

IBEIS (Image Based Ecological Information Systems) is a project that gives us the capability of uniquely identifying individuals (currently limited to animals with stripes/wrinkles/any distinguishing unique pattern) from photographs. IBEIS API's have the capability of extracting different ecological features as well as biological features from the photograph. For more information on IBEIS refer IBEIS home. The IBEIS API's forms the basis of all feature extractions for images in this project.

End Goal:

An intelligent and near accurate population prediction model from Social media photographs is the end goal of this research project. A fully implemented model will scrape public albums of safaris that are shared in various social media platforms. These scraped images will then be fed to IBEIS image recognition system and then feature extraction can be done using IBEIS APIs. Using the extracted features and the learnt population estimation model, population of a certain species limited to a certain geographic region can be then estimated. The proposed estimated models will incorporate self-reporting or behavioral biases into account while predicting population.

Bias Considerations:

The population estimation model should take into consideration the bias of human photographers while performing population estimation. Without considering the bias, the population estimation could be widely off the actual numbers.
There are multiple levels where bias is introduced and we will take into consideration biases introduced starting from the factors that influence the likelihood of a photograph of a certain individual being taken until the same has been shared in a social media platform. The project in its initial phase will try to quantify and recognize any human bias that is introduced when a certain photograph is shared on a public album in a social media. Initial experiments for simulating the sharing of animals on social media is done using an Amazon Web Service tool - Amazon Mechanical Turk (MTurk Wiki).
Read more about Bias consideration here.