H^2A^2：面向统一室内 3D 目标检测的同质感知与异质感知特征提取
H^2A^2: Homogeneity-Aware and Heterogeneity-Aware Feature Perception for Unified Indoor 3D Object Detection
针对室内 3D 检测的跨场景训练，提出结构特征感知的核选择与梯度归一化策略，可复用于点云检测器的跨域泛化与多任务优化。
原文依据 we propose H^2A^2, a homogeneity-aware and heterogeneity-aware feature perception network for unified indoor 3D object detection under cross-scene training paradigms.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论本文观察到，在室内 3D 目标检测中，基本几何线索会在不同场景间引发同质的空间响应，而场景特定结构则产生异质的特征签名。然而，现有检测器缺乏有效机制来联合提取和利用这种双重特性，从而对检测性能造成固有局限。基于这一洞察，我们提出 H^2A^2，一种同质感知与异质感知的特征提取网络，用于跨场景训练范式下的统一室内 3D 目标检测。技术上，我们引入结构特征感知的核选择（SF-KS）方法，包含三个核心组件：（i）任务感知线性调制，一种通道级仿射变换，用于增强场景结构特征表示；（ii）核权重选择策略，集成偏移有效性先验以抑制无信息的跨场景迁移，同时利用结构一致性后验来捕获场景同质线索；（iii）任务感知通道门控，用于抑制场景无关的特征响应。总体而言，SF-KS 能够精确优化同质特征，同时专门处理场景特定的异质特征。此外，为稳定跨场景优化，我们进一步引入基于范数的梯度同质化（NGH）算法，该算法对每个任务的梯度范数进行归一化并动态重加权，以缓解冲突并促进一致更新。在多个室内基准上的大量实验表明，H^2A^2 在强基线上取得一致增益，并提升了跨场景泛化能力。
英文原文摘要 1653 字符
In this work, we observe that for indoor 3D object detection, fundamental geometric cues induce homogeneous spatial responses across scenes, whereas scene-specific structure yields heterogeneous signatures. However, existing detectors lack effective mechanisms to jointly extract and exploit such dual properties, which imposes inherent limitations on detection performance. Guided by this insight, we propose H^2A^2, a homogeneity-aware and heterogeneity-aware feature perception network for unified indoor 3D object detection under cross-scene training paradigms.Technically, we introduce a structural-feature-aware kernel selection (SF-KS) method, which encompasses three core components:(i) task-aware linear modulation, a channel-wise affine transformation that strengthens scene-structural feature representation; (ii) kernel weight selection strategy that integrates an offset validity prior to suppress non-informative cross-scene transfer while utilizing a structural consistency posterior to capture scene-homogeneous cues. and (iii) task-aware channel gating that suppresses scene-irrelevant feature responses. Overall, SF-KS enables the precise optimization of homogeneous features while specializing in scene-specific heterogeneous ones. In addition, to stabilize cross-scene optimization, we further introduce norm-based gradient homogenization (NGH) algorithm, which normalizes and dynamically reweights per-task gradient norms to mitigate conflicts and promote consistent updates. Extensive experiments on diverse indoor benchmarks show that H^2A^2 delivers consistent gains over strong baselines and improves cross-scene generalization.