Date: Thursday, November 7, 2024, 10:00-13:00, Cyprus time
Venue: This training event is held as a hybrid event. You are welcome to join us at the Andreas Mouskos Auditorium, José Mariano Gago Hall, The Cyprus Institute. Otherwise please, connect to our live stream of the discussion, available on Zoom (Password: VsSCz1)
Language: English
Registration: Registration for this event is open until Monday, November 4, 2024. Registration form here.
Prerequisites:
Attendees must have prior experience with Python, as all hands-on exercises will be conducted using this programming language. Familiarity with using a terminal or command line interface is also beneficial.
Requirements:
On-site attendees should bring their own laptop to participate in the hands-on practicum.
Agenda
10:00 - 10:40
Assoc. Prof. Mihalis Nicolaou
Title: Large Scale Machine Learning for Visual and Multimodal Data
This talk provides an overview of recent advancements in large-scale ML, touching upon generative modelling, recently released foundation models, and commonly employed architectures. The talk will discuss capabilities, limitations, and ongoing research efforts (e.g., in the context of explainable and efficient architectures), primarily in the context of visual data and tasks.
10:40 - 11:40
Mr. Marios Constantinou
Title: Exploring intricate vision tasks with Grounded SAM2
During the presentation we will cover the various tasks we can accomplish with Grounded SAM2. Grounded SAM2 is the successor of Grounded SAM, which integrates Grounding Dino as an open-set object detector, with the original Segment Anything Model (SAM). This combination enables the user to perform object detection and segmentation of a region, based on text inputs. Some applications that can be performed using Grounded SAM2 are automatic image annotation, image editing using stable diffusion, human motion analysis, and more. After the presentation, there will be a hands-on section where we will explore the model itself.
Coffee Break
12:00 - 13:00
Dr. Kyriaki Kylili
Title: Multimodal Models for Real-World Applications
Discover how scalable machine learning is transforming computer vision tasks through the use of multimodal models. This presentation showcases the 4M: Massively Multimodal Masked Modeling framework and its innovative approach to integrating multiple data modalities for computer vision tasks. Participants will engage in a hands-on session to apply the model in practical scenarios.