Construction image recognition model

A machine learning model for classifying road construction images to six working stages: gravel, asphalt, excavation, sewer-pipe, cabels, geotextile. The model is based on a convolutional neural network (CNN) and is trained on a dataset of 3,000 images. The base model is a pre-trained ConvNeXt model, which is fine-tuned on the construction images. The following hyperparameters are optimized with a grid search: model size, data augmentation, number of trainable layers, (whether to use) weighted sampling and learning rate.