Papers/2609.30395
🧪 Test?View on arXiv

CSCWD: Cross-Scale Channel-wise Knowledge Distillation for Lightweight Tiny Object Detection on Edge Devices

Author1, Author2, Author3, Author4, Author5

knowledge distillationtiny object detectionedge computingreal-time processing
2609.30395
Builder Relevance
80%
1h ago

Abstract

This study presents a framework for improving tiny object detection in aerial imagery by transferring high-resolution spatial representations from a teacher model to a compact student model.

Reality Card

Core Claim

The YOLO11n-CSCWD model achieves a mean average precision of 50.17% at an intersection-over-union threshold of 0.5, improving upon the baseline by 2.92 percentage points.

Method / Result

Achieved 50.32% mAP@0.5 on Raspberry Pi 5 with 82.32 ms latency.

Limitations

The study may face reproducibility concerns due to the specific hardware and configurations used for evaluation.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers