Table 2 Data augmentation techniques used for balancing and generalization.
From: Real time fire and smoke detection using vision transformers and spatiotemporal learning
Augmentation technique | Application purpose | Description |
|---|---|---|
Random rotation | Balancing & generalization | Rotates images randomly by ± 15°–30° to simulate varying orientations of fire and non-fire scenes. |
Random horizontal and vertical flip | Balancing & generalization | Flips images along both axes to replicate different camera perspectives and viewpoints. |
Random cropping | Balancing | Crops specific regions of images to simulate zoom effects and focus variability in fire and non-fire contexts. |
Brightness, contrast, and saturation adjustments | Balancing & generalization | Modifies lighting and color intensity to mimic variations such as fog, smoke haze, sunlight, or artificial illumination. |
