FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras (1707.09476v2)

Abstract: In this paper, we develop deep spatio-temporal neural networks to sequentially count vehicles from low quality videos captured by city cameras (citycams). Citycam videos have low resolution, low frame rate, high occlusion and large perspective, making most existing methods lose their efficacy. To overcome limitations of existing methods and incorporate the temporal information of traffic video, we design a novel FCN-rLSTM network to jointly estimate vehicle density and vehicle count by connecting fully convolutional neural networks (FCN) with long short term memory networks (LSTM) in a residual learning fashion. Such design leverages the strengths of FCN for pixel-level prediction and the strengths of LSTM for learning complex temporal dynamics. The residual learning connection reformulates the vehicle count regression as learning residual functions with reference to the sum of densities in each frame, which significantly accelerates the training of networks. To preserve feature map resolution, we propose a Hyper-Atrous combination to integrate atrous convolution in FCN and combine feature maps of different convolution layers. FCN-rLSTM enables refined feature representation and a novel end-to-end trainable mapping from pixels to vehicle count. We extensively evaluated the proposed method on different counting tasks with three datasets, with experimental results demonstrating their effectiveness and robustness. In particular, FCN-rLSTM reduces the mean absolute error (MAE) from 5.31 to 4.21 on TRANCOS, and reduces the MAE from 2.74 to 1.53 on WebCamT. Training process is accelerated by 5 times on average.

• The paper details the comprehensive LaTeX author guidelines for preparing and submitting manuscripts to ICCV proceedings.
• Key specifications cover manuscript length, formatting intricacies, blind review considerations, and the use of mathematical expressions and citations.
• Adherence to these guidelines ensures manuscript standardization, facilitates the blind review process, and improves the clarity and presentation of research findings.

Examination of \LaTeX\ Author Guidelines for ICCV Proceedings

The document under review is a comprehensive guide for authors submitting papers to ICCV proceedings, specifically focusing on formatting and submission requirements using \LaTeX. The guidelines are detailed, addressing the critical aspects of manuscript preparation to ensure consistency and adherence to the conference standards. Researchers familiar with academic writing and \LaTeX will appreciate the structured approach this document offers.

Abstract and Introduction Format

The abstract is to be formatted with exact specifications, including italicized text, font size, and placement, emphasizing adherence to previous ICCV styles. This ensures abstracts are concise yet informative and fit within stipulated character limits. The introduction provides a roadmap for compliance, noting essential distinctions in the latest version of guidelines, signifying that it supersedes earlier recommendations.

Manuscript Specifications

Key specifications for manuscript submission include language requirements, paper length, and formatting intricacies. The dual submission policy underscores the conference's stance on manuscript originality and exclusivity. Modern updates to paper length restrictions necessitate author diligence to avoid rejection without review, highlighting the importance of compliance with formatting details such as text area dimensions, column width, and margin settings.

Use of Ruler in Reviewing

A novel feature, the printed ruler, is meant to enhance reviewer comments' accuracy, facilitating easier reference to specific manuscript sections despite its complexity in alignment. This specification illustrates an attempt to streamline the review process without compromising document aesthetics in the final submission.

Blind Review Considerations

The guidelines discuss anonymization for blind review, clarifying common misconceptions like authors needing to omit self-citations. The nuanced distinction assists researchers in maintaining the integrity of their scholarly work while complying with anonymity standards.

Mathematical and Citation Formats

Mathematical expressions necessitate numbering, supporting references that facilitate scholarly discussion. Citation conventions follow a structured format to conserve space and ensure alignment with academic standards. Such meticulousness suggests a balanced approach to showcasing research findings while respecting ICCV’s concise documentation traditions.

Formatting and Style

The document specifies type-styles, margins, and page numbering, ensuring the uniformity of published manuscripts. The precise guidelines for sections, headings, and footnotes aim at maintaining academic rigor while accommodating creative expression within the bounds of specified typographical standards.

Graphical Elements and Their Integration

Graphs and illustrations must be centered and scaled effectively, promoting readability in both electronic and print formats. This consideration underscores the document's practical approach to visual data representation, reflecting an understanding of how research is consumed across different mediums.

Implications

While the guidelines primarily focus on technical aspects, they indirectly influence the quality and impact of submitted research through stringent structural requirements. Adherence to these details ensures that papers not only meet technical standards but also convey ideas with clarity and precision, potentially influencing AI developments discussed at ICCV. Future considerations in AI documentation may further refine these guidelines, integrating evolving technological standards with academic traditions.

In conclusion, the \LaTeX\ Author Guidelines for ICCV Proceedings highlight the critical aspects of manuscript preparation and submission. By ensuring standardization and compliance, these guidelines contribute to the advancement of academic discourse in computer vision and associated research fields.