H2OT:为高效的视频姿势转换器提供分层的沙时钟标记器.
IEEE transactions on pattern analysis and machine intelligence
|September 10, 2025
概括
我们介绍了"等级沙令牌器" (H2OT),一个框架,通过修剪冗余的令牌,有效地从视频中估计3D人体姿势. 这种方法显著降低了基于变压器的模型的计算成本,使它们对资源有限的设备实用.
科学领域:
- 计算机视觉 计算机视觉
- 人工智能的人工智能
- 机器学习 机器学习
背景情况:
- 变压器对于基于视频的3D人体姿势估计是有效的.
- 当前视频姿势变压器 (VPT) 的高计算成本限制了它们在资源有限的设备上使用.
研究的目的:
- 开发一种高效的,可插入和使用的框架,用于基于变压器的3D人体姿势从视频中估计.
- 解决现有 VPT 模型的计算效率低下问题.
主要方法:
- 拟议的等级沙令牌器 (H2OT) 框架与令牌修剪模块 (TPM) 和令牌恢复模块 (TRM).
- 从冗余的视频中动态选择代表性的姿势令牌.
- TRM从选定的令牌中恢复时空信息,用于全长序列推断.
主要成果:
- 通过减少中间变压器块的姿势令牌,H2OT显著提高了模型效率.
- 该方法是通用的,与各种VPT模型和管道 (seq2seq,seq2frame) 兼容.
- 在基准数据集上的3D人体姿势估计中实现了高效率和准确性.
结论:
- 保持完整的姿势序列是不必要的,以高效和准确的3D人类姿势估计.
- 在资源有限的设备上,H2OT为高效的基于变压器的视频姿势估计提供了一个实用的解决方案.
相关概念视频
Transformers
1.7K
A device that transforms voltages from one value to another using induction is called a transformer. A transformer consists of two separate coils, or windings, wrapped around the same soft iron core. However, they are electrically insulated from each other.
The iron core has a substantial relative permeability. Therefore, the magnetic field lines generated due to the current in one winding are almost entirely confined within the core, such that the same magnetic flux permeates each turn of both...
The iron core has a substantial relative permeability. Therefore, the magnetic field lines generated due to the current in one winding are almost entirely confined within the core, such that the same magnetic flux permeates each turn of both...
1.7K
Source Transformation
11.1K
Source transformation is a fundamental technique employed in circuit analysis, offering a valuable tool for simplifying complex electrical circuits. This technique involves the replacement of either a voltage source in series with a resistor by a current source in parallel with a resistor, or vice versa. The key concept here is that when the original sources are deactivated (turned off), the equivalent resistance at the circuit's end terminals remains the same.
It is essential to note that when...
It is essential to note that when...
11.1K
Relative Motion Analysis using Rotating Axes
883
Consider a component AB undergoing a linear motion. Along with a linear motion, point B also rotates around point A. To comprehend this complex movement, position vectors for both points A and B are established using a stationary reference frame.
However, to express the relative position of point B relative to point A, an additional frame of reference, denoted as x'y', is necessary. This additional frame not only translates but also rotates relative to the fixed frame, making it...
However, to express the relative position of point B relative to point A, an additional frame of reference, denoted as x'y', is necessary. This additional frame not only translates but also rotates relative to the fixed frame, making it...
883
Transformers with Off-Nominal Turns Ratios
523
In scenarios involving parallel transformers with disparate ratings, developing per-unit models requires accommodating off-nominal turns ratios. This situation arises when the selected base voltages are not proportional to the transformer’s voltage ratings. Consider a transformer where the rated voltages are related by the term a. If the chosen voltage bases satisfy a relationship involving term b, term c is defined as the ratio of these bases. This ratio is then substituted into the...
523
Transformation
711
Microbial communities are dynamic environments where cell lysis releases free DNA into the surroundings. Other cells can take up this extracellular DNA through a process known as transformation.When a cell incorporates this foreign DNA into its genome, resulting in genetic modification, the process is known as transformation. Cells capable of this process are termed competent. Competence can be natural, as observed in certain bacteria and archaea, or artificially induced in the...
711
Transformations of Functions III
175
Transformations modify the graphical representation of a function without changing its fundamental form. One common transformation is reflection, which flips the graph across a designated axis. When the vertical coordinates of all points are multiplied by the negative one, the entire graph is mirrored over the horizontal axis. This transformation reverses the vertical orientation of peaks and troughs, akin to signal inversion in electrical systems, where a waveform is flipped, but the timing of...
175

