CVPR 18 - Stanford Vision & Learning Lab
The First Workshop on Joint Detection, Tracking, and Prediction in the Wild
Trajectory Forecasting Challenge - CVPR 2018.
Introduction
This CVPR workshop is focused on visual human motion understanding. Identifying where people are in a scene (People Detection), how they are moving (People Tracking), and where they will go to in the future (Motion Prediction) is essential for a number of applications. For example, safe mobile robot navigation (including autonomous driving) depends on robots (cars) being able to recognize where nearby pedestrians are and what they might do next. Likewise, human motion understanding is key for smart surveillance, athletic performance analysis, and VR applications, among others.

This workshop is a full-day event to be held in conjunction with CVPR 18. In this workshop, we want to bring together the computer vision community to envision ways to advance the state-of-the-art in visual human motion understanding. Due to the complexity of this problem, we believe that we should build bridges to unite the communities working on a wide range of interconnected tasks, such as people detection, tracking, and motion prediction. For example, tracking can provide information about the past motion of individuals and, thus, help trajectory forecasting approaches. At the same time, motion prediction can aid in the design of intelligent tracking systems. Moreover, people detection provides fundamental information for both tracking and prediction. But in the presence of occlusions in complex environments, tracking and prediction can help estimate where people might be located next. Our workshop is trying to help solve these different tasks, which are all important, open problems in computer vision.
Workshop Goals
1
Trajectory Forecasting Challenge
2
Bringing Communities Together
3
Shaping Future Research
Call for Papers
Visual human motion understanding is a key computer vision task that aims at understanding the placement, behavior, and future actions of humans in a scene. This field includes the key tasks of finding people in a scene through people detection, understanding movement through people tracking, and predicting future people trajectories through motion prediction. Recently, there has been increasing interest from the academic vision community about this topic, as well as from communities in industry due to its applications in a number of fields.

For example, safe mobile robot navigation, including autonomous driving depends on robots (cars) being able to recognize where nearby pedestrians are and what they might do next. Likewise, human motion understanding is key for smart surveillance, athletic performance analysis, and VR applications, among others.

This workshop aims to provide an overview of the current research trends and advances being made in this field. We also hope to offer academic and industry researchers alike the opportunity to interact, in the hopes that by exposing workshop attendees to different points of view, we can consider to what degree concepts or approaches from one problem area can help advance other related tasks.

Papers addressing topics related to visual human motion understanding are invited. Specifically, we are looking for paper submissions in the fields of people detection, tracking, and motion prediction to our workshop. TrajNet 2018 also has the associated challenge of submitting a motion prediction algorithm on TrajNet. We have made three challenges available: World H-H TRAJ, which looks at predicting motions in world plane coordinates of human-human interactions, Image H-H TRAJ, which looks at predicting the human-human interactions in image plane coordinates, and Image H-H-S TRAJ, which looks at predicting the human-human-space interactions in image plane coordinates. Authors of the top four submissions will have an opportunity to quickly showcase their work as spotlight talks during the workshop. All accepted papers will be presented as posters.
Important Dates
Dec 20, 2017
Web site released.
April 5, 2018
Submission deadline.
April 10, 2018
Submissions reviewed.
April 15, 2018
Acceptances notified.
April 20, 2018
Camera ready date.
June 10, 2018
Challenge deadline.
June 22, 2018
Date of workshop.
Tentative Schedule
Start Time End Time Description
10:00 PST 10:15 PST Opening Remarks
10:15 PST 11:00 PST Invited Talk (30 min talk / 15 min Q&A, Discussion) [Larry Davis]
11:00 PST 11:45 PST Invited Talk (30 min talk / 15 min Q&A, Discussion) [Yaser Sheikh]
11:45 PST 12:00 PST Spotlight: Paper 1 | Paper 2 (4 min talk / 2 min Q&A, Discussion)
12:00 PST 13:30 PST Lunch Break
13:30 PST 14:15 PST Invited Talk (30 min talk / 15 min Q&A, Discussion) [Rita Cucchiara]
14:15 PST 15:00 PST Invited Talk (30 min talk / 15 min Q&A, Discussion) [Farshid Mousavi]
15:00 PST 16:00 PST Poster Session & Coffee Break
16:00 PST 16:45 PST Invited Talk (30 min talk / 15 min Q&A, Discussion) [Patrick Lucey]
16:45 PST 17:45 PST Panel (Alexandre Alahi (EPFL), Patrick Lucey (STATS), Farshid Mousavi (Samsung))
17:45 PST 18:00 PST Closing Remarks
Speakers
Yaser Sheikh
Larry Davis
Patrick Lucey
Rita Cucchiara
Farshid Mousavi
Carnegie Mellon University
University of Maryland
Director of AI at STATS
Univ. of Modena & Reggio Emilia
Director at Samsung
Submission Instructions
Format and Policy
Your submission must be written in English and must be sent in PDF format. Each submitted paper must be no longer than 8 pages excluding references. Please refer to the CVPR author submission guidelines for instructions regarding formatting, templates, and policies, The CVPR submission guidelines can be found at http://cvpr2018.thecvf.com/submission/main_conference/author_guidelines.

By submitting to this workshop, authors agree to the review process and understand that we will do our best to match papers to the best possible reviewers. When we are reviewing, the process will be double blind, in that authors won't know the names of reviewers, and reviewers won't know the names of authors.
Submission Site
Submissions can be made at https://cmt3.research.microsoft.com/trajnet2018/Submission/Index. Note that the submission website requires cookies and Javascript to be enabled. Please also add cmt@microsoft.com to your email provider's whitelist to ensure the submission site's announcements are not marked as spam. If you have any questions about submitting, please contact us at amirabs@stanford.edu.
Organizers
Sponsors
Samsung Toyota Research Institute