Table 5.

Comparison of the baseline and ESS-MB trained on House100K on the room classification task for images from the House14K and Apt14K datasets

Model	Classification dataset	Training loss ↓	Test loss ↓	Test accuracy (%) ↑
Baseline	House14K	0.19 ± 0.003	0.19 ± 0.004	98.10 ± 0.08
ESS-MB	House14K	0.08 ± 0.002	0.08 ± 0.002	99.35 ± 0.08
Baseline	House14KLighting	0.88 ± 0.01	0.93 ± 0.01	78.70 ± 0.85
ESS-MB	House14KLighting	0.47 ± 0.02	0.52 ± 0.03	87.37 ± 0.54
Baseline	Apt14K	1.30 ± 0.04	1.30 ± 0.03	74.85 ± 0.26
ESS-MB	Apt14K	0.65 ± 0.03	0.64 ± 0.03	89.84 ± 0.97