README.md

# IIIF Illustration Detection
a simple cnn for the binary classification into "contains illustration"/"doesn't contain illustration"
directories for training should be structured like this:

train <br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ill<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;noill<br>
    
val<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ill<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;noill<br>
    
test<br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;test


        
also contains a notebook that takes a barcode and gives predictions for pages

getpredictions_ex get barcodes from metadata and predicts for all of them (takes a long time)
saves number of found images into a new column of the metadata-dataframe (to inspect the results)