How to perform a search over millions of points considering multiple image parts in a linear and scalable way?
Here’s the latest result of our Artificial Intelligence Lab. ( Recod.ai ) at Universidade Estadual de Campinas in partnership with the University of Notre Dame and Oak Ridge National Laboratory in the U.S. A groundbreaking paper in the IEEE Transactions on Image Processing (T-IP) representing an important push forward for Multiple Objects Image Retrieval.
Images from social media can reflect diverse viewpoints, heated arguments, and expressions of creativity, adding new complexity to retrieval tasks. We propose a new approach that aims at modeling object-level regions using image parts without the need for costly object detection steps. We call this method the Objects in Scene to Objects in Scene (OS2OS) score, and it is optimized for fast matrix operations, which can run quickly on either CPUs or GPUs.
Explicative video: https://lnkd.in/eyYkxka
Paper with source-code: https://lnkd.in/e66YaN2
I hope you enjoy it!
Thanks, Joel Brogan for all the hard work! 👏😍
Clique abaixo para acessar