Fig. 1. C2f module

Fig. 2. YOLOv8_ZN network structure

Fig. 3. Analysis of bounding box regression characteristics

Fig. 4. EMA module

Fig. 5. Space and channel reconstruction pyramid module

Fig. 6. Pictures of common defects in cylindrical honeycomb ceramics. (a) Ceramic physical object; (b) lateral crack; (c) stuff; (d) scratch

Fig. 7. Data augmentation. (a) Rotation; (b) flip over; (c) random cropping; (d) histogram equalization

Fig. 8. Comparison of small crack identification before and after improvement. (a) Recognition effect of original model; (b) improved model recognition effect

Fig. 9. Comparison of results before and after improvement