Table 14 Image augmentations and their parameters
From: DECIMER 1.0: deep learning for chemical image recognition using transformers
| Image augmentations | Parameters (imgaug) |
|---|---|
| Gaussian blur | 0–1.8 |
| Average blur | k = 0–3 |
| Additive Gaussian noise | Scale = (0, 0.1 * 255) |
| Salt and pepper | 0–0.05 |
| Salt | 0–0.05 |
| Pepper | 0–0.05 |
| Coarse dropout | 0–0.01, size percent = 0.9 |
| Gamma contrast | 0.5–2.0 |
| Sharpen | Alpha = (0.0, 1.0), lightness = 1.0 |
| Enhance brightness | Factor = (0.95, 1.5) |