Vukasin Bozic

Research on 4D reconstruction and perception — recovering geometry and motion of dynamic scenes over time, and the perception models built on top of them.

Doctoral studies at the Photogrammetry and Remote Sensing group, on projects across computer vision, 3D vision, and deep learning. Current work covers panoramic geometry foundation models (PaGeR), 3D mesh simplification and optimization, and diffusion models.

Deployed and improved state-of-the-art monocular depth estimation models for 3D scene reconstruction in stereo conversion pipelines, adapting flat 2D scenes and movies for 3D and VR viewing.

My master thesis project was published at ACM SIGGRAPH 2024. The work exploits large pre-trained text-to-image diffusion models for image and video colorization, achieving state-of-the-art quality with highly controllable, efficient, and customizable output.

Automated brand and product curation for an e-commerce business. Built cloud functions on Google Cloud Platform interacting with Firestore and Storage, and designed data processing and comprehension pipelines driven by generative AI via Firebase Genkit plugins and prompt engineering. Developed and maintained Genkit plugins for OpenAI, Groq, Anthropic, and Mistral.

Taught Probabilistic Artificial Intelligence and Introduction to Machine Learning. Designed ML and AI student projects in which students implemented recent methods in probabilistic AI, and improved the Docker-based cross-platform automatic evaluation and submission interface.

Contributed to research and development of a specialized PTZ camera head for video production automation. Designed automated tracking algorithms for precise PTZ and PTU control as part of my bachelor thesis, and implemented an IMU-based gyro-stabilization algorithm to keep the head stable across diverse scenarios. Projects ranged from low-level microcontroller programming to deep-learning-based tracking.

Photogrammetry and Remote Sensing.

Information Technology and Electrical Engineering. GPA 5.6 / 6.0.

Signal Processing and Control. GPA 9.8 / 10.0.