A TF.Keras Python Implementation
This is a simple implementation of the TableNet nerual net.
It is implemented as a python module that can be imported for training and usage.
Absolutely no guarantee that it works... or warranty if it doesnt.
It is trained on a customized Marmot dataset.
This code has largely been based on the work of others (ex. https://github.com/jainammm/TableNet) but has been largely rewritten and expanded on to make it more generalized.
The initial marmot set used to train this NN is included here in a multi-part zip. It includes 509 images that have been anotated in the Marmot format. This includes column definition data.
Used in PataBid Quantify.