I’d like to propose adding support for inbuilt data loading of image-based datasets in PyDeepFlow, starting with the MNIST dataset. Currently, the datasets.py file includes built-in loaders for tabular datasets like Iris, Wine, and UCI OptDigits. Extending this to cover popular image datasets like MNIST would be a natural and valuable enhancement.
This addition will allow users to directly load MNIST for experimentation and benchmarking without manual downloads, improving usability and aligning with the package’s transfer learning capabilities.
Looking forward to your feedback on this proposal.
Thanks.