Text-to-Video Retrieval For Searching Among a Database of Short Videos

Student: Kirillova Anastasiia

Educational Programme: Math of Machine Learning (Master)

Year of Graduation: 2024

Text-to-Video Retrieval is a task of ranking videos in a database based on their correspondence to the given text query. Recent methods is based on CLIP model, adapt contrastive loss for image and text embeddings alignments, and use additional losses for better multi-modal intergrity. In this work we train an algorithm for a database of short videos. The augmentation pipeline was widely explored, and better video encoder initialisation strategy was proposed. The optimal configuration of implementation details were chosen and the efficiency of the proposed method was set as a primary interest. The analysis of train datasets was conducted and the model was trained on the best set of datasets. The trained model is tested on several benchmarks and compared with other algorithms.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses