Unsupervised Person Re-identification: Clustering and Fine-tuning (1705.10444v2)

Abstract: The superiority of deeply learned pedestrian representations has been reported in very recent literature of person re-identification (re-ID). In this paper, we consider the more pragmatic issue of learning a deep feature with no or only a few labels. We propose a progressive unsupervised learning (PUL) method to transfer pretrained deep representations to unseen domains. Our method is easy to implement and can be viewed as an effective baseline for unsupervised re-ID feature learning. Specifically, PUL iterates between 1) pedestrian clustering and 2) fine-tuning of the convolutional neural network (CNN) to improve the original model trained on the irrelevant labeled dataset. Since the clustering results can be very noisy, we add a selection operation between the clustering and fine-tuning. At the beginning when the model is weak, CNN is fine-tuned on a small amount of reliable examples which locate near to cluster centroids in the feature space. As the model becomes stronger in subsequent iterations, more images are being adaptively selected as CNN training samples. Progressively, pedestrian clustering and the CNN model are improved simultaneously until algorithm convergence. This process is naturally formulated as self-paced learning. We then point out promising directions that may lead to further improvement. Extensive experiments on three large-scale re-ID datasets demonstrate that PUL outputs discriminative features that improve the re-ID accuracy.

• The paper introduces an unsupervised framework that leverages clustering to generate pseudo-labels for person re-identification.
• It employs a fine-tuning strategy to refine deep feature representations and boost re-ID accuracy.
• Empirical results demonstrate that the approach outperforms existing baselines on standard re-ID benchmarks.

Overview of "Bare Advanced Demo of IEEEtran.cls for IEEE Computer Society Journals"

The paper "Bare Advanced Demo of IEEEtran.cls for IEEE Computer Society Journals" by Michael Shell and colleagues focuses on the utilization of the IEEEtran.cls class file in LaTeX for preparing papers intended for IEEE Computer Society journals. This document serves as a comprehensive guide for authors who wish to adhere to the IEEE journal format when using LaTeX, providing a baseline for creating standardized documents.

Purpose and Structure

The primary goal of this paper is to assist authors in the initial setup and formatting of their manuscripts according to IEEE guidelines. It provides an illustrative example of how to implement the IEEEtran.cls class for achieving a professionally styled document suitable for IEEE Computer Society publications.

The paper is organized into several sections, including a template for essential components such as the abstract, keywords, introduction, main content, conclusion, appendices, acknowledgments, and references. It adopts a clear structure that follows the architectural requirements typically expected in IEEE journals:

• Title and Author Block: Demonstrates the correct positioning and formatting of the title and author information, including affiliations.
• Abstract and Keywords: Illustrates the layout and structuring of these critical components for effective dissemination in databases.
• Section Headings: Provides guidelines for organizing content into hierarchical sections and subsections.
• Conclusion and Appendices: Concludes the document with exemplary sections for wrapping up the discussion and including supplementary information.
• Acknowledgments and References: Displays the method for crediting contributors and citing previous work according to IEEE standards.

Technical Details

The guide acknowledges both the technical requirements and document aesthetic considerations pivotal in academic publishing within the IEEE framework. Considerations include:

• Hyphenation and Justification: Ensures proper typographic alignment and spacing.
• Bibliographic Entries: Aligns with IEEE reference styles involving the use of citation commands for consistency.
• Author Affiliations and Contributor Notes: Details placement strategy to affix the necessary affiliations and acknowledge significant contributions properly.

Implications and Future Directions

While this document primarily serves as a useful tool for authors preparing manuscripts for review and publication, it also touches on the broader implications of using standardized typesetting practices. The consistent use of such templates facilitates easier peer review, indexing, and retrieval processes, making scholarly communication more efficient.

Future developments could explore the expansion of the template capabilities to support evolving IEEE format requirements or integrate tools that enhance collaborative manuscript preparation. Moreover, there's potential for advancing interoperability with other document preparation systems or platforms to streamline submission workflows further.

In conclusion, the paper provides an essential resource for authors engaging with IEEE manuscript preparation, ensuring adherence to publication standards and contributing to the seamless dissemination of scholarly work.