Context

As a big Simpsons fan, I have watched a lot (and still watching) of The Simpson episodes -multiple times each- over the years. I wanted to build a neural network which can recognize characters. Feel free to check and recommend my Medium post Part 1 on a classification model and Part 2 on a detection model (Faster R-CNN) about this dataset and what I am doing with it. You can also find the related GitHub repo here .

Content

I am still building this dataset (labeling pictures), I will upload new versions of this dataset. Please check the files there are descriptions and explanations.

File simpson-set.tar.gz : This is an image dataset: 20 folders (one for each character) with 400-2000 pictures in each folder.

File simpson-test-set.zip. : Preview of the image dataset

File weights.best.h5 : Weights computed, in order to predict in Kernels.

File annotation.txt : Annotation file for bounding boxes for each character