CatFLW | tech4animals

CatFLW: Cat Facial Landmarks in the Wild Dataset

Facial landmarks are crucial for a wide range of tasks, including facial analysis, automated detection of pain and affective computing. In contrast to the human domain, the number of datasets containing facial landmarks of animals is exceptionally limited. The ones that do exist usually contain a relatively small number of landmarks, which does not allow for analyzing complex morphological features of animal faces.

The Tech4Animals Lab is proud to introduce the first of its kind Cat Facial Landmarks in the Wild (CatFLW). The dataset contains 2016 images of cats' faces in various environments and conditions, annotated with 48 facial landmarks and a bounding box on the cat’s face. The used images are a subset from the introduced cats dataset in Zhang et al.

THE LANDMARKS

The 48 landmarks were introduced in Finka et al , and were specifically chosen for their relationship with underlying musculature, and relevance to cat-specific facial Action Units (catFACS). In Feighelstein et al they have been shown effective for accurate automated recognition of pain from facial images.

ear left top

ear left lower middle

ear left upper intersection with head

eye l upper

pupil l inner

pupil l upper

pupil l lower

ear right upper middle

ear right top

ear right upper intersection with head

ear right lower middle

ear right lower intersection with head

eye r upper

eye r outer

pupil r outer

pupil r lower

eye r lower

nose l upper

nose r upper

whiskerspad r upper

nose r ventral outer

whiskerspad r bet

whiskerspad r inner

whiskerspad r mid

whiskerspad r outer

pupil r upper

pupil r inner

eye r lower

ear left upper middle

eye l outer

pupil l outer

eye l lower

eye l inner

nose l upper

whiskerspad l upper

whiskerspad l bet

whiskerspad l outer

whiskerspad l mid

whiskerspad l inner

lip l corner

lip l mid

lip lower middle

chin

lip r corner

lip upper middle

ear left lower intersection with head

lip r mid

nose bottom

This unprecedentedly large number of landmarks enables the analysis of complex movements of facial muscles. We believe that this dataset will be a valuable resource for studying the internal state of cats and other cat-related tasks through developing computer vision models for cat facial landmarks' detection.

The dataset is created using a semi-supervised method of annotating images with landmarks, which reduces annotation time and can be used for creating similar datasets for other animals. We also propose universal metrics for evaluating the accuracy of computer vision models on datasets containing animal facial landmarks. Please see CatFLW: Cat Facial Landmarks in the Wild Dataset paper for details.

We firmly believe that animal affective computing crucially depends on the availability of openly available annotated datasets. We are proud to lay the foundations for automated analysis of cat facial expressions, and we invite the scientific community to help us develop this direction. Please contact us if you want to get involved in this research, or request access to the dataset.

The dataset creation was generously supported by the Data Science Research Center (DSRC) of the University of Haifa.

Project Team