Table 4. Comparison of predicted scores and ground truth scores for various sets of images from the Isola et al. Dataset.

Comparison of different networks with the proposed one with another dataset.

Dataset	MemNet	MCDRNet	EMNet	ResMem-Net	Ground Truth
Top 10	80.16%	81.75%	82.43%	82.61%	96.54%
Top 25	75.46%	79.77%	81.41%	82.14%	94.39%
Top 50	75.13%	78.57%	79.48%	80.27%	92.24%
Top 100	74.32%	76.64%	77.63%	79.76%	89.59%
Top 200	73.58%	74.9%	76.64%	77.92%	85.33%
Bottom 200	35.91%	34.83%	34.08%	35.95%	22.85%
Bottom 100	32.8%	31.64%	31.66%	31.4%	18.66%
Bottom 50	30.14%	29.41%	28.94%	29.67%	14.93%
Bottom 25	28.81%	26.6%	29.47%	29.23%	10.95%
Bottom 10	28.29%	26.52%	26.86%	27.42%	5.69%