GSA-YOLO enables high-efficiency real-time X-ray security inspection via structured sparsity and adaptive knowledge distillation
Jiahao Kong · 2026
Read GSA-YOLO enables high-efficiency real-time X-ray security inspection via structured sparsity and adaptive knowledge distillation by Jiahao Kong (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13473521.txt.
Description
Jiahao Kong proposed in 2026 a faster, lighter detector for prohibited items in X-ray security screening. Airport and checkpoint scanners must find hidden objects through heavy overlap and clutter, in real time, yet existing models trade accuracy against speed. The new framework, GSA-YOLO, builds on the small YOLOv8n detector and adds three parts. Group lasso sparsity in the network neck gives robust features. Sparse structure selection in the detection head slims the model, and adaptive knowledge distillation recovers accuracy lost to pruning.
Tests on two X-ray datasets, HiXray and PIDray, gave a leading speed of about 190 frames per second. Computing cost trimmed from 8.7 to 8 gigaflops. Accuracy at the standard mAP50:95 measure reached 0.531 on HiXray and 0.679 on PIDray, improving on the baseline by 2.4 percent and by 1.8 percent in turn. The authors present the model as a practical balance of speed and accuracy for live security lanes.
Catalog record
| CREATOR | Jiahao Kong |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |