Metadata-Version: 2.1
Name: mnistusps
Version: 0.1.0
Summary: Revised splits for MNIST-USPS domain adaptation experiments
Home-page: https://github.com/LukasHedegaard/mnist-usps
Author: Lukas Hedegaard
License: UNKNOWN
Description: # Splits for MNIST-USPS domain adaptation experiments
        
        This repository contains the revised split protocol for creating splits for few shot domain adaptation on the MNIST-USPS datasets.
        
        Contrary to often seen splits, we define an independent test split here and only let the train-val split vary according to a user-defined random seed.
        
        ## Installation
        ```bash
        pip install mnist-usps
        ```
        
        ## Usage
        Getting the splits is a simple as:
        
        ```python
        from mnistusps import mnistusps
        
        train, val, test = mnistusps(
            source_name = "mnist",
            target_name = "usps",
            seed=1,
            num_source_per_class=200,
            num_target_per_class=3,
            same_to_diff_class_ratio=3,
            image_resize=(240, 240),
            group_in_out=True, # groups data: ((img_s, img_t), (lbl_s, _lbl_t))
            framework_conversion="tensorflow",
            data_path = None, # downloads to "~/data" per default
        )
        ```
        
        The function automatically downloads and unpacks the data using Torchvision internally. It then creates the splits using the [Dataset Ops library](https://github.com/LukasHedegaard/datasetops). 
        Depending on your choice of machine learning library, the dataset can be converted to Tensorflow or PyTorch (assuming either is pre-installed) using Dataset Ops.
        
        
        
Platform: UNKNOWN
Requires-Python: >=3.6
Description-Content-Type: text/markdown
Provides-Extra: tests
Provides-Extra: build
