Ekaterina Glazkova, Yandex, head of the VLM product applications team.
The speaker has worked in computer vision for about 7 years (HSE and ShAD, taught at HSE, the HSE Deep Learning and Bayesian Methods Centre, Kaspersky Lab). Yandex's computer vision service works on analysing images and video, and now also jointly with text: discriminative models (embedders, detectors, classifiers), generative models (image and video generation, image modification), and text recognition (OCR). Currently a lot of attention goes to multimodality.
Talk plan: what multimodality and VLMs are and how they are built; why and how to adapt a VLM to product tasks; concrete product applications of Yandex VLM.
Google or a one-time email code. No password.
