Dataset filtering

[jlynxdev]

Description

I created a dataset filtering script which consists of two parts.

• The first part computes the dataset's average mean pixel intensity and average standard deviation per image. It removes the image files whose mean pixel intensity or standard deviation differs from the average over a specified threshold. This method is very good at detecting images that are completely or nearly black or "garbage".
• The second part uses the PCA algorithm to compress and reconstruct image data in order to detect outliers. It effectively identified unnatural looking images or hand-drawn pictures.

This PR also contains the updated dataset with outlier images removed.

[jlynxdev]

Changes introduced

I added a new duplicate image filter. I also carried out a general refactoring of the code so now it's better structured. I introduced separate filter classes and a general compose for stacking filters together. I also restored the dataset to its original form for reference and testing purposes.

[jlynxdev]

Changes introduced

• boolean parameters have been added to extract() methods to introduce optionality of decorators
• the filtered image files have been removed from the dataset folder
• the filter-data.py module has been renamed to data_filtering.py

The module seems to be in a state now where it can be merged.