Automated Cat Facial Analysis Using AI
Author Information
Author(s): Martvel George, Lazebnik Teddy, Feighelstein Marcelo, Meller Sebastian, Shimshoni Ilan, Finka Lauren, Luna Stelio P. L., Mills Daniel S., Volk Holger A., Zamansky Anna
Primary Institution: University of Haifa
Hypothesis
Can automated landmark detection improve the analysis of cat facial expressions for pain and breed recognition?
Conclusion
Automated landmark-based approaches show promise for assessing pain and breed recognition in cats, although performance varies by task.
Supporting Evidence
- Automated pipelines achieved 75% accuracy in cephalic type recognition and 66% in pain recognition.
- The study utilized a dataset of 1,662 images annotated across 18 common cat breeds.
- Manual landmarking was found to be labor-intensive, taking over 5.5 minutes per image.
Takeaway
This study used computers to help understand how cats show their feelings on their faces, which can help us know when they are in pain or what breed they are.
Methodology
The study developed AI pipelines for automated recognition of cat breed, cephalic type, and pain using two datasets of cat faces.
Potential Biases
The reliance on existing datasets may introduce bias due to unbalanced representation of breeds and cephalic types.
Limitations
The automated detection system showed reduced accuracy compared to manual landmarking, particularly in breed recognition.
Participant Demographics
The study included images of cats from 18 common breeds and 29 domestic short-haired female cats undergoing surgery.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website