## Pokemon images for one shot learning One-shot-learning is a very interesting task. This blog has a good [explanation for one-shot-learning](https://sorenbouma.github.io/blog/oneshot/). ## Dataset description There are three directories, "pokemon-a" and "pokemon-b" are standard pokemon images with no background. One can use them for training. "pokemon-tcg-images" are pokemon images cropped from pokemon tcg card. They are for testing. Filenames in all directories start with pokemon-id, aka class label. ## Challenge A classic deep convolution network with softmax can only achieve accuracy below 30%. If you have any good ideas to get better result on this dataset, you are welcome to share with us. Have fun! ## Where those images from All images are collected from internet, and **original authors owns the copyright** (Nintendo, I guess). Check links below for more information: - [kaggle-pokemon-images](https://www.kaggle.com/kvpratama/pokemon-images-dataset) - [veekun](https://veekun.com/dex/downloads) - [pokemon-tcg](https://github.com/PokemonTCG)