RetinaTrack: Online Single Stage Joint Detection and Tracking (2003.13870v1)

Abstract: Traditionally multi-object tracking and object detection are performed using separate systems with most prior works focusing exclusively on one of these aspects over the other. Tracking systems clearly benefit from having access to accurate detections, however and there is ample evidence in literature that detectors can benefit from tracking which, for example, can help to smooth predictions over time. In this paper we focus on the tracking-by-detection paradigm for autonomous driving where both tasks are mission critical. We propose a conceptually simple and efficient joint model of detection and tracking, called RetinaTrack, which modifies the popular single stage RetinaNet approach such that it is amenable to instance-level embedding training. We show, via evaluations on the Waymo Open Dataset, that we outperform a recent state of the art tracking algorithm while requiring significantly less computation. We believe that our simple yet effective approach can serve as a strong baseline for future work in this area.

• The paper provides comprehensive LaTeX formatting and submission guidelines, including page limits (8 excl. references) and English language requirement, for authors submitting to CVPR.
• It details the blind review process protocol, requiring authors to fully anonymize submissions, including citations to their own prior work.
• Adherence to these guidelines is essential for authors submitting to CVPR to ensure successful blind peer review, maintain formatting uniformity, and facilitate inclusion in proceedings.

Overview of \LaTeX\ Author Guidelines for CVPR Proceedings

The paper "\LaTeX\ Author Guidelines for CVPR Proceedings" serves as an essential resource for authors intending to submit their manuscripts to the IEEE Computer Society Press, specifically for the CVPR conference. It provides comprehensive instructions regarding the formatting and submission requirements necessary for preparation and acceptance into the conference proceedings.

Major Components of the Guidelines

1. Language and Submission Policies: The manuscript must adhere to language standards, being exclusively authored in English. The guidelines also delineate the policy on dual submissions, ensuring authors recognize the importance of original and unique contributions to avoid duplicate publishing.
2. Page Limitations and Formatting: The document strictly enforces a limit of eight pages, excluding references. Papers exceeding this limit will not be reviewed, emphasizing the significance of adhering to formatting guidelines, including margin and font specifications. This ensures uniformity and readability across the conference proceedings.
3. Blind Review Process: The protocol for anonymizing submissions is pivotal to the review process. Authors are instructed to avoid personal pronouns like "my" or "our" in order to maintain anonymity while citing previous work, thus ensuring an unbiased review.
4. Mathematical Notations: It underscores the necessity for numbering both sections and equations to facilitate easier referencing for readers and reviewers. The paper references external material for best practices in writing mathematically dense sections.
5. Visuals and References: The treatment of illustrations, graphs, and photographs is addressed to ensure that figures are effectively integrated and informative even when printed. The set standards for references ensure proper citation practices aligned with academic norms.

Implications and Future Considerations

This paper primarily serves an administrative role, structuring the submission norms that culminate in the proceedings of the CVPR conference. Its implications are broadly practical rather than theoretical, ensuring that submissions meet a standardized format for coherent and accessible presentation of ideas.

For authors and researchers in computer vision, strict adherence to these guidelines is vital for facilitating the peer review process while ensuring the graphical and textual coherence of their work. The guidelines mitigate potential biases in the review phase by maintaining the anonymity of submissions.

Looking forward, such guidelines are likely to evolve alongside advancements in digital publishing technologies and methods of scholarly communication. The increasing importance of open science and data accessibility may also influence future iterations of the guidelines, potentially integrating more comprehensive frameworks for the inclusion of supplemental data and interactive figures. Additionally, as digital manuscript preparation evolves, formats might adapt to incorporate machine-readable elements for AI-driven analysis and indexing within scholarly databases.

In summary, this paper is instrumental in both guiding authors through procedural requirements and supporting the integrity and uniformity of academic publishing in the field of computer vision research, fostering a robust and equitable platform for knowledge dissemination.