These models were trained to (a) classify, (b) reconstruct, and (c) inpaint the
PaRCE Park Dataset. For each of these models, layers.json is a JSON file that defines the model architecture, train.config is a configuration file that contains all of the training/optimization parameters, and model.pth is a serialized PyTorch state dictionary storing the learned parameters of the trained model. For the reconstruction and inpainting models, there is also a pickle file called parce.p, which contains the trained Probabilistic and Reconstruction-based Competency Estimation (PaRCE) model.
