Context The data was originally collected from here: https://www.kaggle.com/c/painter-by-numbers but only train_2.zip was available on new Kernels. So the following datasets can be combined to hold all the needed data in 1 new Kernel: * [painter-test](https://www.kaggle.com/mfekadu/painter-test/) * [painters-train-part-1](https://www.kaggle.com/mfekadu/painters-train-part-1/) * [painters-train-part-2](https://www.kaggle.com/mfekadu/painters-train-part-2/) * [painters-train-part-3](https://www.kaggle.com/mfekadu/painters-train-part-3/) # This is a messy way to get around Kaggle's 20 GB dataset limit, but you can [**_just fork this notebook_**](https://www.kaggle.com/mfekadu/painter-by-numbers-combined-dataset) to quickly get started. Content test.zip Acknowledgements Thank you [Kiri Nichol](https://www.kaggle.com/smallyellowduck) for collecting the data for the competition. Inspiration Do the pixels that represent a Picasso painting, uniquely identify him?