CSCWD: Cross-Scale Channel-wise Knowledge Distillation for Lightweight Tiny Object Detection on Edge Devices
CSCWD: Cross-Scale Channel-wise Knowledge Distillation for Lightweight Tiny Object Detection on Edge Devices
CSCWD：用于边缘设备轻量级微小目标检测的跨尺度通道知识蒸馏
Abstract: Real-time tiny object detection in aerial imagery is constrained by the weak spatial evidence of very small objects and the loss of high-resolution detail in lightweight detectors. This study presents Cross-Scale Channel-wise Knowledge Distillation (CSCWD), a training-time framework that transfers high-resolution spatial representations from a YOLO11m-P2 teacher to a compact YOLO11n student without altering the student’s inference architecture.
摘要： 航空影像中的实时微小目标检测受到微小目标空间特征微弱以及轻量级检测器高分辨率细节丢失的限制。本研究提出了跨尺度通道知识蒸馏（CSCWD），这是一种训练时框架，它将高分辨率空间表征从 YOLO11m-P2 教师模型迁移到紧凑的 YOLO11n 学生模型，且无需改变学生模型的推理架构。
Unlike conventional same-scale feature distillation, CSCWD transfers supervision from teacher P2 to student P3 after feature alignment while retaining same-scale distillation at deeper pyramid levels. Under the unified seven-sequence Drone-vs-Bird validation protocol, YOLO11n-CSCWD achieves 50.17% mean average precision at an intersection-over-union threshold of 0.5 (mAP@0.5) and 59.73% recall, improving the matched CA-YOLO11n baseline by 2.92 percentage points in mAP@0.5 and 3.55 points in recall.
与传统的同尺度特征蒸馏不同，CSCWD 在特征对齐后将监督信息从教师模型的 P2 层迁移到学生模型的 P3 层，同时在更深的金字塔层级保留同尺度蒸馏。在统一的七序列 Drone-vs-Bird 验证协议下，YOLO11n-CSCWD 在交并比（IoU）阈值为 0.5 时达到了 50.17% 的平均精度均值（mAP@0.5）和 59.73% 的召回率，相较于匹配的 CA-YOLO11n 基准模型，mAP@0.5 提升了 2.92 个百分点，召回率提升了 3.55 个百分点。
Cross-scale alignment further increases mAP@0.5 by 2.09 points over the corresponding same-scale channel-wise distillation configuration. In zero-shot evaluation on DUT-Anti-UAV, mAP@0.5 increases from 48.29% to 50.06% without target-domain fine-tuning. This domain was included because its challenging small targets make low-latency, computationally efficient detection particularly relevant.
跨尺度对齐使 mAP@0.5 在相应的同尺度通道蒸馏配置基础上进一步提升了 2.09 个百分点。在 DUT-Anti-UAV 的零样本评估中，无需目标域微调，mAP@0.5 从 48.29% 提升至 50.06%。引入该领域是因为其具有挑战性的微小目标使得低延迟、计算高效的检测显得尤为重要。
On Raspberry Pi 5 using NCNN-FP16 at 640x640 resolution, the 2.58-million-parameter student achieves 50.32% mAP@0.5 at 82.32 ms mean wall-clock latency, or 12.15 frames per second, while retaining essentially the same runtime and memory requirements as the matched baseline. The results support cross-scale distillation for improving tiny-target detection without increasing inference-time model complexity.
在树莓派 5 上使用 NCNN-FP16、640x640 分辨率进行测试时，该拥有 258 万参数的学生模型在 82.32 毫秒的平均实际延迟（即每秒 12.15 帧）下实现了 50.32% 的 mAP@0.5，同时保持了与匹配基准模型基本相同的运行时间和内存需求。研究结果表明，跨尺度蒸馏能够在不增加推理阶段模型复杂度的情况下，有效提升微小目标检测性能。