InvoiceNet is a deep learning system for extracting structured information from invoice documents, published on GitHub with about 2,700 stars. It classifies invoice fields and pulls values such as totals, dates and invoice numbers out of PDF and image invoices automatically.
InvoiceNet combines document layout analysis with neural text classification: it segments an invoice into text blocks, classifies each block into field types, and applies extraction rules to produce clean, structured output. A graphical interface lets users view invoices, inspect extracted fields and correct them, while a Trainer UI supports building custom models on your own invoice dataset and adding or removing fields to match your document formats. The project is MIT licensed and built on Keras and TensorFlow. The codebase was last updated in 2024. Sample invoices, data preparation scripts and documentation for extending the field extractor are included, making it a common starting point for student and startup document-AI projects.
InvoiceNet is completely free and open source under the MIT license. There are no paid tiers or hosted versions; you train and run it on your own machines.
InvoiceNet is for finance teams, bookkeeping services and developers who process large volumes of invoices and want to automate data entry from them.