Forecasting model of an apartment interior quality assessment

The entire code and project was written in the Python programming language Imports libraries such as Numpy, CV2, PyTorch, Albumentations, Pandas, Atexit Using pandas, we set settings to make the display window output, giving the maximum number of rows, columns. Using time, datetime, atexit we create a function to measure the time when the program started, when the program was completed, how long it was used. By announcing the variable train, we read a training data set with the photo name and their assessment. After that, we use OneHotEncoder to convert the ratings into a more extensive look. Next, create a class TestDataset, to process the photos that are stored in the folder on the server, specify the path of the folder, describe the transformation for augmentation, using CV2 we open each photo, change its size to 224x224, after transforming and saving in a pixel. With albumentations, we transform the photo and store it in tensor. Then we read the test data set, which will be tested, model. With torch.utils.data.DataLoader, we load our test dataset. Then load our pre-trained model (based on Resnet50). We convert the output through tangential function. Keep the score in a separate column and each class in different columns. After each class, we translate into a conditional coefficient, for a better understanding of the results of the model.