From its very foundation, Mirum has been working on a brand new class of recognition algorithms for the retail field. Our goal is to drastically reduce the need for big data structures in training computer vision AI. Just like a human can recognize an item after looking at a photo once, so should a machine be able to do so in the near future. Our goal is to make sure our algorithms learn to recognize objects as fast as humans do.
In order to stimulate the development of this brand new class of machine vision algorithms, called "one-shot-detection", we're sharing some of our knowledge in this field in a new scientific article called "OS2D: One-Stage One-Shot Object Detection by Matching Anchor Features
" published on arXiv.org.
Our investigations in this area of computer vision allow Mirum to provide high recognition accuracy and advanced quality of our solutions for retailers and brands.
The paper is available at https://arxiv.org/abs/2003.06800