Our workshop is happening on June 17，2024. This is a hybrid workshop. Check out our program below.

Workshop Goals

Welcome to the 5th workshop in the JRDB series! Our workshops are designed to tackle perceptual problems that arise when autonomous robots operate and interact in human environments including human detection, tracking, forecasting, and body pose estimation, as well as social grouping and activity recognition.

In this workshop, we are excited to explore the challenging task of Open-Word Panoptic Segmentation and Tracking in human-centred environments. We will introduce a new competition that challenges participants to develop models capable of accurately predicting panoptic segmentation and tracking results using the JRDB dataset. We have also invited speakers from the field of visual perception and robotics to offer valuable insights into understanding human-centred scenes.

Call for Papers

Guidelines

We invite researchers to submit their papers addressing topics related to autonomous robot in human environments. Relevant topics include, but not limited to:

2D or 3D human detection and tracking
2D or 3D semantic, instance or panoptic segmentation
2D or 3D open-world recognition
2D or 3D skeleton pose estimation and tracking
Human trajectory forecasting
Human motion prediction and safety
Visual scene prediction
Human-robot Interaction considering predictions
Visual and social navigation in crowded scenes
Human motion/body skeleton pose prediction
Predictive planning and control
Individual, group and social activity recognition
Human walking behaviour analysis
Dataset proposals and bias analysis
New metrics and performance measure for different visual perception problems related to autonomous robot

Timeline

Submission deadline for the full papers:
April 4 - 23:59 PT
Acceptance notification of full papers:
April 12 - 23:59 PT
Camera-ready deadline for the full papers:
April 14 - 23:59 PT
Submission deadline for the extended abstracts:
June 5 - 23:59 PT
Acceptance notification of the extended abstracts:
June 17 - 23:59 PT

More information

The Best Paper Award will be granted if more than five papers are accepted.

Submissions

Full papers are up to eight pages, including figures and tables, in the CVPR style. Additional pages containing only cited references are allowed. For more information, please refer to the guidelines provided by CVPR here. Extended abstracts should also adhere to the CVPR style and are restricted to a maximum of one page, excluding references. Accepted papers have the opportunity to be presented as a poster during the workshop. However, only full-papers will appear in the proceedings. By submitting to this workshop, the authors agree to the review process and understand that we will do our best to match papers to the best possible reviewers. The reviewing process is double-blind. Submission to the challenge is independent of the paper submission, but we encourage the authors to submit to one of the challenges.

Submissions website: here. If you have any questions about submitting, please contact us here.

Submit to our Workshop

Accepted Papers

InViG: Benchmarking Open-Ended Interactive Visual Grounding with 500K Dialogues

Hanbo Zhang, Jie Xu, Yuchen Mo, Tao Kong

Must Unsupervised Continual Learning Relies on Previous Information?

Haoyang Cheng, Haitao Wen, Heqian Qiu, Lanxiao Wang, Minjian Zhang, Hongliang Li

HumanFormer: Human-centric Prompting Multi-modal Perception Transformer for Referring Crowd Detection

Heqian Qiu, Lanxiao Wang, Taijin Zhao, Fanman Meng, Hongliang Li

GM-DETR: Generalized Multispectral DEtection TRansformer with Efficient Fusion Encoder for Visible-Infrared Detection

Yiming Xiao, Fanman Meng, Qingbo Wu, Linfeng Xu, Mingzhou He, Hongliang Li

Pre-trained Bidirectional Dynamic Memory Network For Long Video Question Answering

Jinmeng Wu, Pengcheng Shu, HanYu Hong, Ma Lei, Ying Zhu, Wang Lei

DSTCFuse: A Method based on Dual-cycled Cross-awareness of Structure Tensor for Semantic Segmentation via Infrared and Visible Image Fusion

XUAN LI, Rongfu Chen, Jie Wang, Ma Lei, Li Cheng

Is Our Continual Learner Reliable? Investigating Its Decision Attribution Stability through SHAP Value Consistency

Yusong Cai, Shimou Ling, Liang Zhang, Lili Pan, Hongliang Li

Open Challenge

We are organizing a new challenge on Closed- and Open-World Panoptic Segmentation and Tracking .

The challenge includes:

Closed-world panoptic segmentation
Open-world panoptic segmentation
Closed-world panoptic tracking
Open-world panoptic tracking

Each challenge submission should be followed by an extended abstract or full paper submission via our paper submission portal (see details above) or a link to an existing preprint/publication. The top 3 will have the opportunity to present their work as a spotlight (5 minutes) in the workshop.

Challenge Guidelines

Participants must strictly adhere to the submission policy provided on the main JRDB webpage, which can be found here
Submission Tag: Each submission must include the tag "_CVPR24" following the submission name, e.g., "YourSubmissionName_CVPR24". Submissions lacking this tag will not be considered.
Submission Details: Along with your submission, please provide:
- Name: Clear and concise name of your submission.
- Abstract: Brief description outlining the key aspects and innovations of your method.
- Paper Link: A link to a detailed paper or preprint (e.g., arXiv) discussing your methodology and results.
Deadline Compliance: Ensure that all required information is complete by the submission deadline. Incomplete submissions will be disqualified post-deadline.

Evaluation Protocol

Submissions will be evaluated based on:

Closed-World Settings: Methods will be evaluated using 42 classes for both panoptic segmentation and tracking.
Open-World Settings: Methods will be evaluated using all 72 classes.

All methods are evaluated on stitched images from the dataset at 1Hz (every 15th images). Participants can also use individual camera views for training, supported by provided camera calibration parameters and stitching code. For more details on the dataset, visit JRDB-PanoTrack dataset.

Leaderboard

The best submission from each account will be displayed automatically during challenge period, encouraging competition and showcasing leading-edge methods.

Timeline

Data release and challenge start: April 1 - 23:59 PT
For dataset information and download:
See JRDB-PanoTrack
For evaluation toolkit:
See JRDB-PanoTrack Toolkit
For data manimulation tutorials:
See Tutorials
Leaderboard available and submissions allowed: April 14 - 23:59 PT
Submit to benchmark:
Submit
Challenge close:
June 1 - 23:59 PT

Time (relative)	Speaker	Title
0:00 - 0:10		Introduction
0:10 - 0:40	Speaker 1
0:40 - 1:00	Organizers	Introducing the JRDB Dataset and the Challenge
1:00 - 1:30	Speaker 2
1:30 - 1:45	Speaker 3
1:45 - 2:15		Coffee break & poster session
2:15 - 2:45	Speaker 4
2:45 - 3:15	Speaker 5
3:15 - 3:30		Challenge winners presentations
3:30 - 4:00	Speaker 6
4:00 - 4:10

Invited Speakers

Rita Cucchiara

Full Professor at University of Modena and Reggio Emilia, Italy

Laura Leal-Taixe

NVIDIA | Adjunct Professor at Technical University of Munich

Deva Ramanan

Professor at Carnegie Mellon University, USA

Dariu Gavrila

Full Professor at Delft University of Technology, the Netherlands

Michael Milford

Professor at Faculty of Engineering, Queensland University of Technology, Australia

Program Committee

Name	Organization
Apoorv Singh	Motional
Gengze Zhou	The University of Adelaide
Haodong Hong	The University of Queensland
Houzhang Fang	Xidian University
Jiarong Guo	Hong Kong University of Science and Technology
Jin Li	Shaanxi Normal University
Michael Wray	University of Bristol
Mingtao Feng	Xidian University
Qi Wu	Shanghai Jiao Tong University
Shuai Guo	Shanghai Jiaotong University
Shun Taguchi	Toyota Central R&D Labs., Inc.
Tiago Rodrigues de Almeida	Örebro University
Weijia Liu	Southeast University
Zijie Wu	University of Western Australia
Ziyu Ma	Hunan University