About

We are witnessing a paradigm shift in computer vision, driven by the explosion of video data from diverse platforms and the rise of multimodal foundation models. The MAVU workshop focuses on multi-faceted video understanding across three distinct modalities: aerial/drone vision, surveillance video, and egocentric vision.

The workshop explores key problems including forensic video search, Video Question Answering (VQA), and long-video understanding. The core theme of this workshop is to promote research into flexible interactions with these diverse video sources. By leveraging natural language queries (VQA) and advanced forensic search capabilities, we can transform passive video archives into interactive intelligence systems. This approach allows users to intuitively navigate and reason about complex events across different scales and viewpoints, bridging the gap between raw pixel data and high-level semantic understanding.

MAVU serves as a premier forum for researchers to present novel methodologies that handle the specific nuances of each modality while pushing the boundaries of interactive video understanding. By fostering discussion on these multi-faceted approaches, we aim to accelerate the development of robust, real-world applications in public safety, autonomous systems, smart cities, and next-generation media intelligence.

Call for Papers

We invite original research contributions in (but not limited to) the following areas:

• Long-video understanding and reasoning

• Video question answering and multimodal reasoning

• Video retrieval and forensic search

• Drone and aerial video analysis

• Surveillance video understanding and human action recognition/crowd analysis

• Egocentric and first-person video understanding

• Aerial video perception, including detection, tracking, and segmentation

• Efficient and real-time video models for UAVs and edge devices

Submissions can include short papers (up to 4 pages including references) or full papers (up to 8 pages excluding references) in the WACV main conference format. Accepted full papers will be included in the WACV proceedings. We welcome papers that propose a new technical approach for any of the above, or claim to take a position regarding challenging and open-ended questions that address multi-facet analysis for aerial and remote sensing video understanding

Important Dates

• Paper Submission Deadline: October 20, 2026 11:59 PM PST

• Decision Notification: October 30, 2026

• Challenge Participation Deadline: November 10, 2026, 11:59 PM PST

• Camera Ready Submission Deadline: November 20, 2026 11:59 PM PST

Submission Preparation Instructions

Please follow the main conference format and submission guidelines to prepare your papers. Check WACV Submission Guidelines here

Submission site

We will use OpenReview for submissions. Link to the submission portal

All authors need to have an OpenReview profile. Please plan ahead as it can take up to two weeks.

AI Challenge: VQA on Drone Videos

The MAVU workshop will host an AI challenge on Video Question Answering (VQA) for drone videos, designed to benchmark recent advances in video understanding, spatiotemporal reasoning, and remote sensing.

What is the Challenge About?

Participants will develop models that answer natural-language questions about high-resolution aerial video sequences captured by drones. The questions require models to understand objects, scenes, events, spatial relationships, and temporal changes across video frames. The goal is to evaluate how well modern video-language models can reason about complex aerial scenes.

Task

Given a drone video and a natural-language multiple-choice question, participants must predict the correct answer. Questions require understanding and reasoning over both spatial and temporal information in aerial scenes.

Details

4,000 question-answer pairs from 1,014 drone videos
Videos collected across 18 cities worldwide
Average video length: ~15 seconds
Baseline model: BIMBA

Evaluation

Prizes sponsored by Qualcomm:
Champion🥇— $1,000
2nd Place🥈— $500
3rd Place🥉— $200
Primary metric: overall multiple-choice accuracy

Example Question

Questions may require reasoning across multiple video frames. For example: “How long does the yellow taxi remain visible in the video frame?”

Workshop Presentation

Top-performing teams will be recognized during the MAVU workshop. Selected teams will be invited to present their solutions and findings at WACV 2027.

Participate in the Challenge

Challenge submissions and leaderboard evaluation are hosted on Codabench.

Go to Codabench →

Keynote Speakers

We are delighted to welcome our keynote speakers to the workshop.

Dinesh Manocha

University of Maryland at College Park

Keynote Talk

Towards General Audio-Visual Intelligence

Xiaoming Liu

University of North Carolina at Chapel Hill

Keynote Talk

Advancing Video Analytic in the Era of Vision Foundation Models

Yang Liu

Qualcomm

Keynote Talk

To be announced

Program Schedule

Date:
Time: 09:00 – 18:30

Time	Event
09:00 – 09:10	Opening Remarks
09:10 – 10:00	Invited Talk 1
10:00 – 10:50	Invited Talk 2
10:50 – 11:10	Coffee Break
11:10 – 12:00	Invited Talk 3
12:00 – 13:30	Lunch Break and Poster Session 1
13:30 – 14:20	Invited Talk 4
14:20 – 15:10	Invited Talk 5
15:10 – 15:40	Coffee Break
15:40 – 16:40	AI Challenge Session: overview, top solutions, and awards
16:40 – 17:30	Poster Session and Industry Demos
17:30 – 18:15	Panel Discussion: “The Future of Video Intelligence: From Forensics to Autonomy”
18:15 – 18:30	Best Paper Awards and Closing Remarks

🌐 Website

Contact

For questions about the workshop, please contact the organizers at mavu-wacv27-organizers@googlegroups.com.