Table 4 The performance with ambiguity rejection precision.
From: Ambiguity-aware semi-supervised learning for leaf disease classification
Data ratio | Base accuracy | Coverage | Select accuracy | Reject accuracy |
|---|---|---|---|---|
BRACOL dataset | ||||
   05% | \(92.18 \pm 1.45\) | \(72.54 \pm 5.05\) | \(99.50 \pm 0.48\) | \(31.58 \pm 4.12\) |
   10% | \(93.94 \pm 0.94\) | \(76.16\pm 2.51\) | \(99.43 \pm 0.44\) | \(23.60 \pm 4.06\) |
   20% | \(95.26 \pm 0.59\) | \(82.00 \pm 3.81\) | \(99.42 \pm 0.41\) | \(24.11 \pm 4.35\) |
   25% | \(95.98 \pm 0.47\) | \(85.88 \pm 2.00\) | \(99.08 \pm 0.61\) | \(22.65 \pm 4.06\) |
   33% | \(96.33 \pm 0.49\) | \(84.03 \pm 4.04\) | \(99.33\pm 0.14\) | \(19.57 \pm 1.12\) |
   50% | \(96.58 \pm 0.69\) | \(88.51 \pm 2.76\) | \(\varvec{99.46 \pm 0.37}\) | \(25.18 \pm 2.96\) |
   100% | \(96.78 \pm 0.68\) | \(86.16 \pm 4.11\) | \(\varvec{99.38 \pm 0.24}\) | \(20.53 \pm 7.01\) |
Banana dataset | ||||
   05% | \(94.09 \pm 2.07\) | \(91.95 \pm 5.23\) | \(98.56 \pm 0.90\) | \(61.72 \pm 14.41\) |
   10% | \(96.87 \pm 1.31\) | \(91.12 \pm 2.61\) | \(99.71 \pm 0.18\) | \(49.91 \pm 12.10\) |
   20% | \(98.26 \pm 0.32\) | \(96.90 \pm 0.44\) | \(99.97 \pm 0.16\) | \(51.64 \pm 6.44\) |
   25% | \(98.76 \pm 0.36\) | \(97.29 \pm 1.46\) | \(99.76 \pm 0.31\) | \(49.61 \pm 33.83\) |
   33% | \(98.76 \pm 0.31\) | \(98.04 \pm 2.34\) | \(99.69\pm 0.31\) | \(77.78\pm 38.49\) |
   50% | \(99.23 \pm 0.22\) | \(97.99\pm 1.10\) | \(\varvec{100.0 \pm 0.00}\) | \(41.67 \pm 11.79\) |
   100% | \(99.31 \pm 0.26\) | \(99.50\pm 0.04\) | \(\varvec{99.69 \pm 0.02}\) | \(63.33 \pm 41.50\) |
