*3.1. Image Database Creation*

A smartphone is used for image acquisition. For the purpose of collecting images of small cracks on a concrete surface, all images are taken with a distance of 0.1 m between the smartphone and the concrete surface. Two-thousand original images with sizes of 3024 × 3024 pixels are extracted from the surfaces of concrete buildings. Each original image can be cropped to generate 139 images with sizes of 256 × 256 pixels. However, several cropped images do not include cracks. As a result, the images with cracks are meticulously selected from the cropped image set. Finally, 10,000 images conforming to the requirements are selected to create the database.

To assess the generalization ability of the improved model, the 10,000 images are divided into five parts according to the fivefold cross-validation principle, of which 80% are used to train and validate the model and the remaining 20% are used to test. More precisely, 8000 images are randomly selected from the 10,000 images, among which 7000 images are used to generate a training set and 1000 images are used to create a validation set. The remaining 2000 images not selected for training or validation are used to build a testing set.
