Table 4 Summary of testing the method over a synthetic dataset with multiple low-temperature heat sources.
From: Tomographic reconstruction from planar thermal imaging using convolutional neural network
| Â | Original model (four-block CNN) | Extended model (five-block CNN) | ||||
|---|---|---|---|---|---|---|
Metrics | Mean ± std. dev. | Median (25–75 percentile) | Mean ± std. dev. | Median (25–75 percentile) | ||
NMSE | 0.022 ± 0.017 | 0.016 | (0.012–0.025) | 0.014 ± 0.017 | 0.008 | (0.007–0.012) |
CC | 0.971 ± 0.020 | 0.977 | (0.968–0.981) | 0.980 ± 0.021 | 0.987 | (0.982–0.989) |
PSNR | 29.579 ± 2.005 | 29.314 | (28.545–30.649) | 32.082 ± 2.852 | 31.834 | (30.487–33.661) |
SSIM | 0.841 ± 0.074 | 0.849 | (0.786–0.900) | 0.904 ± 0.040 | 0.910 | (0.879–0.938) |
