Projects

We introduce Cat Facial Landmarks in the Wild (CatFLW), a dataset of cat images annotated with 48 facial landmarks. This dataset can aid the development of computer vision models and studying cats' internal state.

Citizen science is becoming more popular in animal behavior research. petsDATAlab is a platform that allows researchers to quickly create mobile apps for data collection through pet owners. This eliminates the need for expert developers and designers and reduces the time required to build the app from months to just a few clicks.

Recognizing emotional states in animals is important, but identifying animal affect is challenging for dogs. This research project presents the first dataset of dog faces in positive and negative emotions, collected through a controlled experiment, to address this gap. The dataset contains dog faces cut from video clips of 29 Labrador Retrievers in two emotional states: Negative (frustration) and Positive (anticipation).

This project aims to develop AI models for estimating dog age, a task that has been overlooked despite the fact that dogs are the most well-studied species in animal science and their ageing processes are similar to those of humans. To achieve this, a large-scale, crowdsourced data collection is being launched, and dog owners are invited to contribute photos of their pets. 

There is a need to develop non-invasive animal-based indicators of affective states in livestock species, such as using vocal indicators. Cows produce LF calls for close distance contacts and HF calls for long distance communication, the latter being associated with negative affective states. We present an open-source dataset in project page.

Recent advances in deep learning have improved our ability to track animal behavior in labs. Applying these techniques to larger animals in open spaces, like dogs, is challenging. We're bridging this gap by using wearable sensors to monitor and analyze dog behavior in various real-world situations.