ACTION-Net: Multipath Excitation for Action Recognition (2103.07372v1)

Abstract: Spatial-temporal, channel-wise, and motion patterns are three complementary and crucial types of information for video action recognition. Conventional 2D CNNs are computationally cheap but cannot catch temporal relationships; 3D CNNs can achieve good performance but are computationally intensive. In this work, we tackle this dilemma by designing a generic and effective module that can be embedded into 2D CNNs. To this end, we propose a spAtio-temporal, Channel and moTion excitatION (ACTION) module consisting of three paths: Spatio-Temporal Excitation (STE) path, Channel Excitation (CE) path, and Motion Excitation (ME) path. The STE path employs one channel 3D convolution to characterize spatio-temporal representation. The CE path adaptively recalibrates channel-wise feature responses by explicitly modeling interdependencies between channels in terms of the temporal aspect. The ME path calculates feature-level temporal differences, which is then utilized to excite motion-sensitive channels. We equip 2D CNNs with the proposed ACTION module to form a simple yet effective ACTION-Net with very limited extra computational cost. ACTION-Net is demonstrated by consistently outperforming 2D CNN counterparts on three backbones (i.e., ResNet-50, MobileNet V2 and BNInception) employing three datasets (i.e., Something-Something V2, Jester, and EgoGesture). Codes are available at \url{https://github.com/V-Sense/ACTION-Net}.

Overview of "LaTeX Guidelines for Author Response"

The document "LaTeX Guidelines for Author Response" provides an explicit framework for authors preparing to submit a rebuttal following the review of their paper for the CVPR conference. This guide is crucial for authors aiming to address reviewer comments effectively while adhering to the submission constraints imposed by the conference organizers.

Key Elements of the Guidelines

The paper outlines several key elements and procedural advisories for crafting author rebuttals:

• Purpose and Scope: The rebuttal serves to clarify factual inaccuracies or provide additional clarifications as requested by reviewers, within the constraints of the original submission. Importantly, it is not an avenue for introducing novel contributions beyond what was initially submitted.
• Length and Format: The rebuttal is strictly confined to a single page, including references and any figures. This limitation necessitates concise and precise communication. The content must be formatted in a specified two-column layout to align with CVPR standards.
• Content Restrictions: Authors are guided not to present new experiments or results not included in the initial submission. This aligns with the 2018 PAMI-TC motion discouraging reviewers from expecting supplementary experiments during the rebuttal phase. Any tables or figures included should reflect data from the original submission or existing literature, not fresh analyses.
• Blind Submission Requirements: The response must maintain the anonymity protocol of the original submission. Authors are advised to update the paper title and ID appropriately while abiding by these stipulations.

Implications and Technical Considerations

Through these guidelines, the document underscores the importance of conformity to structured academic dialogue and standardization in submissions. Such uniformity aids in maintaining fairness and efficiency in the review process while avoiding the pitfalls of subjective interpretation or inconsistency in evaluating rebuttals.

These guidelines, though seemingly procedural, hold significant implications for academic discourse. Firstly, they ensure that the focus remains on the quality and clarity of arguments rather than sheer volume or supplemental data that reviewers may not have initially requested. This also indirectly accentuates the need for authors to preemptively consider reviewers' potential queries during the submission phase itself, enhancing the overall rigor and quality of their research.

Future Considerations

Looking forward, as AI and machine learning continue to evolve, the guidelines could be supplemented with automated tools for ensuring compliance or even initial reviewer feedback assessment, reducing the cognitive load on human reviewers. Furthermore, enhanced natural language processing techniques might offer authors preemptive insights into possible rebuttal areas based on early reviews, fostering proactive refinement of their research narratives.

These developments might not only streamline the rebuttal process but could also enhance the robustness of scientific communications, providing a template for advancing peer-review systems in computer science conferences and beyond.