Adam Kortylewski

Title: Adam Kortylewski and Occlusion-Aware Object Detection

Introduction

Adam Kortylewski is an inventor associated with Johns Hopkins University in Baltimore, Maryland. His patent work addresses a challenge in computer vision: detecting objects in images when parts of them are hidden by other objects.

Latest Patents

Kortylewski is associated with 1 patent: “Systems, Methods, and Computer Programs for Using a Network of Machine Learning Models to Detect an Image Depicting an Object of Interest Which Can Be Partially Occluded by Another Object.”

The patent describes methods and systems for detecting objects in images despite partial occlusion. A deep convolutional neural network processes image features, which are then analyzed by an occlusion model and a generative compositional model. The occlusion model identifies regions where an object is obscured, while the compositional model detects object classes corresponding to parts or components. Their outputs contribute to a likelihood map indicating whether the object of interest appears in the image.

Career Highlights

Kortylewski’s listed invention applies machine learning to object recognition, combining visual context, object features, and compositional information to improve detection when an image is incomplete or obstructed. The work is associated with Johns Hopkins University.

Collaborations

Alan Yuille is identified as Kortylewski’s coworker. Their shared research setting at Johns Hopkins University connects their work to computer vision and machine learning.

Conclusion

Adam Kortylewski’s patent presents an approach to more robust object detection by combining neural-network features with models that account for occlusion and object composition.