If you’ve ever had a pet cat, you know they are masters of communication. Cats were domesticated over 10,000 years ago, learning to mix body language with an assortment of meows, purrs, and chirps to express their needs to humans. I also have a cat colony with complex social relationships based on factors like rank, age, sexual status, and genetics. So how do they “speak” to each other?
Researchers have indicated that other mammals, including primates, communicate using facial signals. For example, gorillas often mirror each other’s facial expressions while playing; this phenomenon is known as rapid imitation of faces. Scientists correlate emotional perception with this rapid facial imitation, which may have evolved as a precursor to human empathy. Veterinarians are particularly aware of how cats exhibit different facial expressions when experiencing fear, irritation, relaxation, or pain. They often display certain behaviors when scared, and lick their noses and hiss when annoyed. However, it’s unclear whether cats in colonies and multi-cat households use imitation of facial signals.
Researchers in Israel and the United States have recently developed a new automated approach to determine if domestic cats utilize facial mimicry. Historically, researchers have compared animal facial expressions by manually recording specific movements based on standardized facial action coding systems.FACS. Other scientists tracked changes in facial shape by observing particular reference points.Landmarks on the faces of animals. Since both methods are time-intensive and subjective, the research team suggested machine learning could expedite the process and reduce bias.
The team analyzed 186 videos of 53 adult short-haired cats. Catcafe Lounge in Los Angeles, California, recorded between August 2021 and June 2022. Related to social interaction, they classified neutral or antagonistic interactions such as staring and hissing as non-intimate interactions. The hypothesis was that cats would mirror each other’s faces more frequently during interactions compared to those who are not acquainted, similar to other mammals that use facial mimicry to bond.
Initially, researchers tested whether machine learning models could accurately classify cat interactions in the videos. They utilized a model known as a tree-based pipeline optimization tool.TPOT, previously used for sorting genetic data. Starting from a manually assembled CATFACS dataset, they tracked 48 different movements involving the lips, ears, and eyes in the videos. They trained TPOT on 147 videos using the CATFACS dataset and tested its accuracy on another 37. The model successfully identified interactions based on the leaders’ facial movements in 74% of the videos.
Following that, the researchers examined how well TPOT characterized cat interactions based on facial landmarks, including 48 reference points covering the eyes, ears, nose, and mouth. They began with anautomatic landmark system that quantified cat facial signals from the video footage. The benefit of automating this approach is that it can capture rapid, subtle movements that humans might overlook. They trained TPOT with over 87,000 video frames using the automated landmark data and tested it on 22,000 frames. They found that TPOT’s ability to classify landmark-based facial signals was not more accurate than that of CATFACS-based signals, suggesting that fully automated landmarks could actually be more prone to error than manual tracking.
Finally, the researchers analyzed the comprehensive CATFACS dataset using TPOT to determine when one cat mimics some or all of another cat’s facial expressions. Supporting their hypothesis, they discovered that cats statistically mimicked each other more during social interactions than during unrelated ones. They also found that cats commonly mimic each other’s ear movements. Their results affirm previous claims that cats use their ears for communication, but they acknowledged that cats might also respond to external sounds instead of each other.
The team concluded that, much like other mammals, cats utilize rapid facial mimics to communicate. They suggested that these facial cues help cats within colonies navigate their intricate social environments and coexist peacefully. However, they also recognized that improving automated landmark-based facial tracking could involve using more cameras or conducting tests in controlled environments to minimize external influences. Regardless, they proposed that automated tracking of cat facial signals could someday enhance the success of living arrangements in shelters and among veterinarians.
Post view: 391
Source: sciworthy.com
