[Retracted] X-Ray Small Target Security Inspection Based on TB-YOLOv5

Figure 7

Reticular formation of four structures. (a) FPN adopts a top-down path to mix different features from P3 to P7. (b) PANet introduces an added down-top path based on FPN. (c) NAS-FPN uses neural networks to exploit unusual information and recycle the same module. (d) BiFPN adds extra branches to attain superior detection accuracy and realize a more efficient trade-off.

(a)

(b)

(c)

(d)