A new method to develop AI model without collecting large amounts of real image data
- Machine learning
- National Institute of Advanced Industrial Science and Technology, Artificial Intelligence Research Center (AIRC)
The National Institute of Advanced Industrial Science and Technology (AIST) has developed the new method for building AI image recognition model (pre-trained model) using a large-scale image data set automatically generated from mathematical equations.
Despite this method does not use real images for pre-training, it achieves image recognition accuracy that is at least as good as conventional methods. It also eliminates the cost of collecting a large amount of image data used for training, the cost of annotation, and the risk of privacy violation. This technology is expected to apply to develop AI in various environments, such as automated driving, medical care, and logistics.
This technology will be presented at the IEEE/CVF International Conference on Computer Vision and Pattern Recognition (CVPR) 2022, to be held in New Orleans, LA from June 19 to 24, 2022.
Please see here , for details. [In Japanese]en