🤖 AI Summary
To address the high computational cost and model complexity in point cloud 3D detection caused by frequent neighborhood searches in multi-scale feature learning, this paper proposes a single-neighborhood multi-scale feature approximation framework. Our method avoids redundant neighborhood construction via knowledge distillation to compress and approximate multi-scale features; introduces a transferable, class-aware statistical embedding mechanism that leverages lightweight class-level statistics to compensate for diversity loss under single-neighborhood sampling; and adopts a center-weighted IoU loss to mitigate localization misalignment induced by center offset. Evaluated on Waymo and KITTI benchmarks, our approach achieves state-of-the-art detection accuracy while reducing inference latency by 2.1× and model parameters by 37%, demonstrating significant efficiency gains without compromising detection performance.
📝 Abstract
This paper investigates multi-scale feature approximation and transferable features for object detection from point clouds. Multi-scale features are critical for object detection from point clouds. However, multi-scale feature learning usually involves multiple neighborhood searches and scale-aware layers, which can hinder efforts to achieve lightweight models and may not be conducive to research constrained by limited computational resources. This paper approximates point-based multi-scale features from a single neighborhood based on knowledge distillation. To compensate for the loss of constructive diversity in a single neighborhood, this paper designs a transferable feature embedding mechanism. Specifically, class-aware statistics are employed as transferable features given the small computational cost. In addition, this paper introduces the central weighted intersection over union for localization to alleviate the misalignment brought by the center offset in optimization. Note that the method presented in this paper saves computational costs. Extensive experiments on public datasets demonstrate the effectiveness of the proposed method.