https://dasya.itu.dk/supervisor/omar-shahbaz-khan/ Recent content in Omar Shahbaz Khan on DASYA Hugo -- gohugo.io en-us © DASYA, IT University of Copenhagen Fri, 18 Sep 2026 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/current/osk-ecpfs-multirepresentation/ Fri, 18 Sep 2026 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/current/osk-ecpfs-multirepresentation/ ecpfs stores one representation of each vector per index. This project adds the ability to store more than one, full vectors, PCA-reduced vectors, or both, and lets you choose which representation gets used at which point in the search: for routing, for final ranking, or both. Once that capability exists, the real question is what combinations are actually worth it: when does keeping a second representation around pay for itself, and when does it just add cost without helping? https://dasya.itu.dk/for-students/proposals/archive/osk-qa-multimedia/ Fri, 21 Nov 2025 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/archive/osk-qa-multimedia/ Constrastive learning models have made it easier to find relevant content from multimedia collections through descriptive text queries, reducing the interactivity needed to solve simple tasks. However, for complex tasks not only pertaining visual elements, or tasks focused on answering questions involving the contents of a singular or multiple media items (videos/images) still requires a fair amount of interactivity. Retrieve results from query Inspect content (summaries, watching, browsing) Refine the query / add more context / different search strategy Repeat until answer to the question or relevant content has been discovered RAG (Retrieval-Augmented-Generation) approaches utilize LLMs (Large-Language Models) to answer questions from a set of documents. https://dasya.itu.dk/for-students/proposals/archive/osk-ecp-bench/ Tue, 14 Jan 2025 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/archive/osk-ecp-bench/ The extended Cluster Pruning (eCP) index, is a hierarchical approximate nearest neighbor index. eCP-FS is a “white-box” implementation of the index as a file structure using zarr. The index building is done through Python, while the index loading and search have been implemented in both pure Python and a Rust-Python package. While the index is slower from disk than other disk-based ANN solutions due to it not being stored as a single or few binary files but tons of smaller files, it allows more control over what is loaded, is easy to extend and when in-memory is on par with other memory based indexes.