This repository contains the source code for neural networks with transcribing capabilities in a Django API. The API is described with Swagger OpenAPI 3.0. See deployment below on how to open the swagger documentation page.
The goal for this project was to learn about the math involved with neural network and how to construct the layers in a neural network. This learning goal was part of my data science minor, where I build a handwriting recognition neural network based on a research paper from Graves A. & Schmidhuber J. called Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks
The goal for this project was to learn about the math involved with neural network and how to construct the layers in a neural network. This learning goal was part of my data science minor, where I build a handwriting recognition neural network based on a research paper from Graves A. & Schmidhuber J. called Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks
- prerequisite - docker installed
- optionally:
- add a keras checkpoint hdf5 file called handwritten_text_model.hdf5 inside src/apps/computer_vision_services_text_reader/files if you do not have this file you will have to use the computer vision train text endpoint to generate the file. This will take a couple of hours. The file is also too large too upload to Github.
- add a keras checkpoint hdf5 file called all_handwritten_characters_model.hdf5 inside src/apps/computer_vision_services_text_reader/files/models
- TODO: split and merge hdf5 files so they can be uploaded to Github
- add a keras checkpoint hdf5 file called handwritten_text_model.hdf5 inside src/apps/computer_vision_services_text_reader/files if you do not have this file you will have to use the computer vision train text endpoint to generate the file. This will take a couple of hours. The file is also too large too upload to Github.
steps
- Run "docker-compose up -d" from the root directory of this project
- Open localhost:8000