O-DRUM @ CVPR 2023

Workshop on Open-Domain Reasoning Under Multi-Modal Settings

June 19, 2023
8:30 AM -- 5:00 PM PDT
Vancouver

ODRUM 2022 Archive: [Webpage] YouTube

AI has undergone a paradigm shift in the past decade -- the connection between vision and language (V+L) is now an integral part of AI, with deep impact beyond vision and NLP -- robotics, graphics, cybersecurity, and HCI are utilizing V+L tools and there are direct industrial implications for software, arts, and media. The link between vision and language is much more complex than simple image--text alignment – the use of language for reasoning beyond the visible (for example, physical reasoning, spatial reasoning, commonsense reasoning, and embodied reasoning) is being pursued. Open-Domain Reasoning in Multi-Modal Settings (ODRUM 2023) provides a platform for discussions on multimodal (vision+language) topics with special emphasis on reasoning capabilities.

The aim of ODRUM 2023 is to address the emerging topic of visual reasoning using multiple modalities (such as text, images, videos, audio, etc.). The workshop will feature invited talks by experts in the realm of reasoning such as: embodied AI, navigation, learning via interaction and collaboration with humans, building large V+L that can perform multiple tasks, visual grounding, and the use of language to instruct robots. Participants and speakers will converge for a panel discussion to discuss the importance of reasoning (a core AI topic that has a rich and long history since the 1950s) to computer vision, relevance to recent progress in visual reasoning, discuss trends and challenges in open-domain reasoning, from different perspectives of NLP, vision, machine learning, and robotics researchers.

Confirmed Speakers

Kristen Grauman
Professor
University of Texas at Austin

Jiajun Wu
Assistant Professor
Stanford University

Alane Suhr
Young Investigator
Allen Institute for AI

Jean_Baptiste Alayrac
Research Scientist
Deepmind

Angel Xuan Chang
Assistant Professor
Simon Fraser University

Tentative Schedule

08:30 -- 08:45	Welcome and Introduction
08:45 -- 09:35	Invited Talk 1
09:35 -- 10:00	Spotlight Talks
10:00 -- 10:45	Poster Session + Coffee Break
10:45 -- 11:35	Invited Talks 2
11:35 -- 12:25	Invited Talk 3
12:25 -- 13:25	Lunch
13:25 -- 14:15	Invited Talk 4
14:15 -- 15:05	Invited Talk 5
15:05 -- 16:00	Poster Session 2 + Coffee Break
16:00 -- 17:15	Panel Discussion + Concluding Remarks

Call for Papers

We invite submissions related to the broad topic area of multi-modal understanding, reasoning and comprehension, including but not limited to following topics:

Visual Reasoning using multiple modalities such as images, videos, audio, and text
Visual Grounding with natural language
Embodied perception, such as vision--language navigation
Use of natural language, instructions, or other vision--language supervision in robotics
Collaborative and multimodal learning, with human interaction and/or feedback
Multimodal information retrieval
Methods of retrieval that facilitate additional tasks, such as image and video captioning, visual grounding, VQA, image generation, and graphics, among others
Utilization of Retrieval in unsupervised/few-shot/zero-shot learning for data augmentation and generation.
New datasets or task designs for open-domain reasoning, use of knowledge bases in multimodal reasoning, or any other multimodal tasks
Modification and enhancement of existing multi-modal comprehension tasks, including VQA, VQA with Knowledge (e.g., OK-VQA, Web-QA, etc).
Design of new evaluation metrics for multimodal tasks
Analysis, commentary, or position pieces on existing evaluation metrics in multimodal tasks
Image and video analytics
Efficient and/or robust representation learning
Reliability and Robustness issues in multimodal learning and reasoning
The role of external knowledge in multimodal reasoning
Continual / Lifelong / Online multimodal learning

We encourage two types of submissions:

Extended abstracts (four pages plus an unlimited number of references)
Long papers (8 pages maximum with unlimited references)

All submitted materials must be anonymized and formatted according to the CVPR 2023 author guidelines and template. Accepted papers will be presented as posters at the workshop, where attendees, invited speakers, and organizers will engage in discussion. We intend to highlight three best papers during the workshop session with spotlight presentations. We will give authors of all accepted papers an option to opt-in or opt-out of CVPR proceedings.

Important Dates:

♦ Submission Deadline:	March 24, 2023, 23:59 PDT
♦ Notification of Decision:	March 31, 2023
♦ Camera Ready Deadline:	April 08, 2022 , 12:00 PDT
♦ Submission website (CMT):	https://cmt3.research.microsoft.com/ODRUM2023/