Technology2022년 09월 07일 2022-09-12 23:09
Automated Extraction & Recognition Technology of Pets Biometrics Information
Real-time detection and classification of companion animalsVarious objects are photographed altogether while recording with a smartphone camera. While photographing dogs and cats, other animals, or your family, may be included. Our in-house developed object detection technology can precisely seek and classify your companion animals. Just dogs and cats can be recognized for registration. (*Beta version of Cat expected to be released in December of this year)
Improvement of individual recognition performance
with artificial intelligence-based quality assessment
All the data cannot be perfect as blurry images are photographed in dark places.
Data of companion animals can be imperfect as they tend to move continuously.
Our technology utilizes proprietary machine learning technology to instantly
assess the quality of photographed nose print or face, and automatically retries
the acquisition process when quality is insufficient.
Higher quality images can be obtained through this process.
Automated camera control
Based on data acquisition experience of biometric information,
Petnow has implemented the most optimal automated
camera control technology.
Our app automatically focuses on a dog's nose or cat’s face,
controls the camera with brightness and quality assessment
result accordingly for obtaining higher quality images.
Minimalization of user effort
After a companion animal is automatically found in the
video while a camera is running in our app, a dog/cat is
categorized, and then the quality of biometrics
information(nose print/face) is inspected.
The camera is controlled accordingly for continuously
improving the quality of acquired data.
With machine learning and artificial intelligence technology,
all the process above is done promptly and automatically.
smartphone, and point at a dog/cat for easily obtaining
high-quality nose print/face images.
Dogs Identification Technology
Artificial Intelligence Deep Neural Network
optimized for dogs identification
For past decades, biometrics-based identification solution has been focused on humans. Attachment of tags, livestock branding, and microchip implantation have been used for identifying animals.
Through our proprietary research technical skills and university
cooperation research, a new artificial intelligence neural network that is optimized for identification of dogs has been developed.
99+% accuracy has been achieved with our technology specialized in nose print identification, and the achievement has been published on IEEE Access successfully after getting approved by peer reviewers.
An AI neural network optimized for
cat’s facial recognition
We designed and trained the AI neural network reflecting cat’s
behavioral traits. Our unique research and development are
tailored just for cats, because their faces are different from humans’.
We attained an impressive recognition rate for cats faces
of up to 99%. This result has been submitted for publication in
international journals as well as for registration as intellectual property.
Assessment of performance
*Same colored points are the same dogs, and performance of differentiation is higher when they are closer.
*Different colored points are different dogs, and performance of differentiation is higher when they are farther.
Google Inception Module shows approx. 90% recognition success rate.
Our deep neural networks shows approx. 98% of recognition success rate.
(as published on IEEE Access: Dog Nose-Print Identification Using Deep Neural Networks. DOI: 10.1109/ACCESS.2021.3068517)
High speed search function
with parallel processing
Millions of nose print information can be inquired instantly
by adapting cutting edgy parallel distributed processing solution
and our neural network designed optimally for looking up
Advancing artificial intelligence deep neural networks
Our artificial intelligence does not stay as is. As more data is collected,
its performance is improved with recurred studies, and
next-gen artificial intelligence neural network technology can be designed.
Our research has been continued to exceed accuracy and convenience of
human biometrics identification technology.
Results of our artifical intelligence dep neural networks consist of
random numbers that are irreversible, nor they can be understood by people.
Once nose print data is processed, it is safely encrypted before
getting stored in a cloud storage.