Application of Partial Differential Equation Image Classification Methods to the Aesthetic Evaluation of Images

Table 1

Comparison of test accuracy between the NP-DP-DCNN method and other methods in a photoquality dataset database: classification 1: animal, classification 2: botany, classification 3: static state, classification 4: architecture, classification 5: scenery, classification 6: character, and classification 7: night view.

Classification	1-DCNN method	2-NP-DCNN method	3-NP-DCNN method	2-NP-DCNN and 3-NP-DCNN combination
1	91.82%	93.07%	93.23%	93.23%
2	87.78%	89.56%	88.40%	89.56%
3	87.66%	89.36%	88.04%	89.36%
4	89.92%	91.76%	92.02%	92.02%
5	92.44%	93.66%	95.00%	95.00%
6	95.58%	94.32%	95.58%	95.58%
7	87.96%	89.66%	88.39%	89.66%
Whole	90.88%	91.91%	91.98%	92.64%
Classification	NP-DP-DCNN method	Literature [7]	Literature [9]	Literature [16]
1	94.33%	78.61%	77.51%	81.61%
2	91.33%	76.38%	80.93%	82.38%
3	89.40%	71.74%	78.29%	81.74%
4	93.11%	73.86%	85.26%	73.86%
5	96.17%	77.53%	81.70%	77.53%
6	96.11%	76.94%	79.08%	77.94%
7	89.56%	64.21%	73.21%	64.21%
Whole	93.24%	74.95%	79.44%	77.92%