Fig. 1. Flow chart of Bayesian Loss model

Fig. 2. Structural diagram of Bayesian Loss model

Fig. 3. Schematic diagram of Bayesian loss function

Fig. 4. Geometric diagram of virtual background point

Fig. 5. Generation of sheep density map. (a) Original drawing ; (b) sheep density map; (c) geometric adaptive Gaussian kernel

Fig. 6. Examples of dataset images. (a) Different illumination; (b) different backgrounds; (c) different number of sheep; (d) different distributions of sheep

Fig. 7. Counting result

Fig. 8. Effect of Gaussian kernel parameter on model. (a) Original image; (b) σ=8; (c) σ=16

Fig. 9. Training loss curves of different models. (a) MCNN; (b) CSRNet; (c) SFANet; (d) Bayesian Loss

Fig. 10. Visualization results of different models