Five sets of two models derived from the digitized images from the twentieth century from the Russian State Documentary Film and Photo Archive (RGAKFD). The models are derived from Inception v3, retrained with TensorFlow. The five sets of models are the following: Decades: Model trained to classify images by decade (1900-1909, 1910-1919 etc.) Twenty Years: Model trained to classify images by twenty year intervals. Fifty Years: Model trained to classify images by fifty year intervals. Political Eras: Model with historically informed intervals (1900-1916, 1917-1928, 1929-1953, 1954-1964, 1965-1984, 1985-1991, 1992-1999, inclusive) Big Political Eras: Model with bigger historically informed intervals (1900-1916, 1917-1940, 1941-1953, 1954-1984, 1985-1999) All files are in a zip folder. The folder contains two Python scripts: 1. A script that compiles an image classification model in TensorFlow with the parameters used in the manuscript. It also gives a history of the accuracy and loss for training and validation data, the classes for the model.2. A Grad-CAM class activization visualization adapted from the sample code at keras.io (https://keras.io/examples/vision/grad_cam/) To run the first script, one needs to have image data sorted into folders and to provide the script with the location of that data. I have not provided the images but have provided the ID numbers of the images I used to compile the models. It also contains ten folders with compiled models and relevant data for each. Each folder contains: 1. The model (NAME.keras)2. A Python dictionary with the prediction classes (classes.dict)3. A Python dictionary listing the training and validation images for each category (data.dict)4. A Python dictionary with the history of accuracy and loss for training and validation for the model compilation (history.dict)5. A Python dictionary with the predictions for each training image (predictions.dict)