InvoiceNet

Why we included this project

InvoiceNet is for teams that still type invoice numbers and totals into a spreadsheet by hand. It offers a desktop UI to view PDF, JPG, or PNG invoices and a neural network that pulls the fields you care about, like vendor, total, and invoice number, off the page, then saves the parsed records into your own system with a click. You are not stuck with whatever the model was trained on: a Trainer UI and data-preparation scripts let you retrain on your own invoice layouts and add or remove fields to match your workflow. The catch is that no general pretrained weights are distributed yet, so plan on labeling a set of your own invoices before extraction gets accurate. If you already have a pile of such documents, this gives you the tooling to turn that pile into a custom extractor rather than continue typing totals by hand.