重新审视词元压缩：加速基于ViT的稀疏多视图3D目标检测器
Revisiting Token Compression for Accelerating ViT-based Sparse Multi-View 3D Object Detectors
把「图像块尺寸」本身当作可动态调节的压缩维度，配合跨粒度特征增强把粗块丢失的细节补回来，可直接迁移到任何 ViT 主干的多视图/时序感知模型上做免训练或轻训练加速。
原文依据 dynamically adjusts patch sizes while preserving critical semantic information within coarse patches
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论基于视觉Transformer（ViT）的稀疏多视图3D目标检测器已取得显著精度，但由于词元处理开销大，推理延迟仍然很高。为加速这类模型，词元压缩已被广泛探索。然而，我们对词元剪枝、词元合并和图像块尺寸增大等现有策略的重新审视表明，它们往往会丢弃有信息量的背景线索、破坏上下文一致性并损失细粒度语义，从而对3D检测产生负面影响。为克服这些局限，我们提出SEPatch3D，一个新框架，它在动态调整图像块尺寸的同时，在粗图像块内保留关键语义信息。具体而言，我们设计了时空感知图像块尺寸选择（SPSS），为包含近处物体的场景分配小图像块以保留细节，为以背景为主的场景分配大图像块以降低计算开销。为进一步缓解潜在的细节损失，有信息量图像块选择（IPS）挑选有信息量的图像块进行特征精炼，跨粒度特征增强（CGFE）则将细粒度细节注入被选中的粗图像块，从而丰富语义特征。在nuScenes和Argoverse 2验证集上的实验表明，SEPatch3D相比StreamPETR基线实现了最高57%的推理加速，相比最优的ToC3D-faster效率高出20%，同时保持相当的检测精度。
英文原文摘要 1461 字符
Vision Transformer (ViT)-based sparse multi-view 3D object detectors have achieved remarkable accuracy but still suffer from high inference latency due to heavy token processing. To accelerate these models, token compression has been widely explored. However, our revisit of existing strategies, such as token pruning, merging, and patch size enlargement, reveals that they often discard informative background cues, disrupt contextual consistency, and lose fine-grained semantics, negatively affecting 3D detection. To overcome these limitations, we propose SEPatch3D, a novel framework that dynamically adjusts patch sizes while preserving critical semantic information within coarse patches. Specifically, we design Spatiotemporal-aware Patch Size Selection (SPSS) that assigns small patches to scenes containing nearby objects to preserve fine details and large patches to background-dominated scenes to reduce computation cost. To further mitigate potential detail loss, Informative Patch Selection (IPS) selects the informative patches for feature refinement, and Cross-Granularity Feature Enhancement (CGFE) injects fine-grained details into selected coarse patches, enriching semantic features. Experiments on the nuScenes and Argoverse 2 validation sets show that SEPatch3D achieves up to 57% faster inference than the StreamPETR baseline and 20% higher efficiency than the state-of-the-art ToC3D-faster, while preserving comparable detection accuracy.