A Deep Neural Network Based on Circular Representation for Target Detection

Figure 6

Statistics of the test results of the trained RebarDet model in 40 rebar pictures, the abscissa is the difference between the predicted results and the real results, and the ordinate is the number of predicted pictures. (a) is the statistical chart of the test results of the RebarDet model with -score of 0.9872 and AB-score of 0.7983, (b) is the statistical chart of the test results of the RebarDet model with -score of 0.9476 and AB-score of 0.8114.

(a)

(b)