Fig. 1. Frame diagram of improved YOLOv5s

Fig. 2. Schematic diagram of attention mechanism

Fig. 3. Frame diagram of channel attention mechanism

Fig. 4. Frame diagram of spatial attention mechanism

Fig. 5. Strengthening of feature fusion

Fig. 6. Dislocation corrosion pits of monocrystalline silicon

Fig. 7. Diagram of model training loss

Fig. 8. Comparison of different network performances

Fig. 9. Effect drawings of acid corrosion pits detected by different models

Fig. 10. Effect drawings of alkaline corrosion pits detected by different models