Panasonic has announced the creation of a new artificial intelligence (AI) algorithm for image recognition.
This algorithm is capable of identifying and tracking a variety of subjects and further categorizing them into specific subcategories based on their visual representation in an image.
Enhanced Object Identification Under Challenging Conditions
The Panasonic X AI system has been designed to recognize objects even when their appearance is altered by factors that usually cause problems for traditional object identification methods. These factors include changes in lighting, orientation, weather, and background.
Improvement in Image Classification Accuracy
Responding to the increasing use of AI in subject recognition across different visual media, Panasonic has utilized a benchmark dataset to develop a technique that significantly improves image classification accuracy. This technique, termed “multimodal distribution,” allows for more precise categorization, such as identifying specific breeds of animals, types of vehicles, or variations in appearance like color.
Solving Traditional AI Recognition Problems
Panasonic has tackled the issue that traditional AI systems face in accurately recognizing objects within the same category that appear differently. The company has developed a new classification algorithm that takes advantage of these differences in appearance. It uses a two-dimensional orthonormal matrix to capture a wide range of image variations, thereby improving the AI’s ability to differentiate between subjects.
Expansion of the Classification Model
To better capture feature distributions, Panasonic has transformed the traditional one-dimensional weight vector of the classification model into a two-dimensional orthonormal matrix. This modification allows each element of the weight matrix to represent different variations in the image, such as changes in background colors or the orientation of objects.
Future Applications and Industry Influence
The timeline for integrating this technology into camera systems is not specified, but Panasonic plans to continue developing this object recognition technology and eventually implement it across various industries. The company is also focused on promoting the research and development of this technology and training AI specialists to further develop and implement it.
Upcoming Conference Presentation
Panasonic is scheduled to present this research at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) in Hawaii, which will take place from January 4 to January 8, 2024. More detailed information about this AI technique is available in a white paper on the WACV website.
News source: https://news.panasonic.com/global/press/en231225-4
