Oron Nir

My research focuses on high-level semantic representation learning and its applications in computer vision. I have a keen interest in developing innovative methods to enhance the understanding and processing of visual data.

My research projects include:
• Character Labeling in Animation using Self-supervision by Tracking: This project suggests a novel method for style-specific animated character representation. The use of multi-object tracking for self-supervised learning improves character labeling in animated videos.

• Video Representation for Contextual Retrieval with LLMs: In this work, I investigate the use of large language models and video understanding models to enhance video representation for more effective contextual search and retrieval in media libraries.

• Vision Language co-embeddings architectures for open-vocabulary applications: Multi-modal architectures bridge language and vision while unlocking traditional supervised methods to cross-modal learning.

• Artistic Style Disentanglement from Content Representation in images: Visual domains like artistic styles significantly influence image analysis and semantic representation. In this work we aim at creating orthogonal representations for content and style.

• Density-based clustering algorithms: High-dimensional representations impose data analysis challenges. Through several of our works we have developed unsupervised methods to extract patterns from noisy unlabeled datasets in high-dimensional spaces.

Besides my academic research work I had also the privilege of working with bright engineers at Microsoft as a co-founder of Azure Video Indexer, developing AI models for video indexing and understanding. Applications in Audio Analysis, Computer Vision, and Natural Language Processing.

Through my research, I strive to contribute to the advancement of computer vision and representation learning, pushing the boundaries of what is possible in these exciting fields. I am constantly looking for collaboration opportunities so, feel free to reach out.