Software Projects/Research

Wild Intelligence Lab – a Software Platform to integrate Computer Vision into Wildlife Management

To enable the usage of the developed data analysis strategies in the working routine of conservation professionals a user-friendly software program needs to be developed. Therefore, we work on integrating the strategies in a modular analysis platform called WildJarvis which combines our three core functions of WildAi, WildAnalsis and WildLabel.

WildAi

Datasets are continuously annotated and consequently increase in size over time. To benefit from this ongoing annotation we want to equip conservation professionals with the ability to train instance segmentation models and to inspect their performance and reliability. The trained weights can be directly used in our function WildAnalysis, allowing WildAnalysis to improve over time through use.

WildAnalysis

The data that can be extracted from the high-resolution drone images through the use of instance segmentation models is vast. This requires a dedicated analysis function that identifies different key data points on the segmentation masks. We will work closely with conservation professionals to develop new ways to analyze the information from the drone images in an attempt to improve the understanding of changing ecosystems. Due to a large amount of data, we strive for a highly automated process that analyses large image datasets automatically, while providing new insights into the nature observed in the images.

WildLabel

Driven by deep learning models, this analysis approach requires large, annotated datasets. The qualitative requirements for the annotations are high, making this task extremely time intensive. To allow the workload to be split among many different data annotators of different domain expertise, we aim to develop a platform that allows anyone to request a dataset for annotation while ensuring the highest annotation quality through inspection by domain experts at Kuzikus Wildlife Reserve.

WildJarvis

Running the developed functions will be very calculation-intensive and requires expensive hardware. Therefore, we aim to find partners to build a powerful hardware platform that can run the WildTech functions to provide Kuzikus Wildlife Reserve with an integrated system that allows our WildTech functions to run smoothly and without internet connection, allowing for the seamless integration into the workflow of conservation professionals.
This website uses cookies to ensure you get the best experience on our website.