4 ms·
Ask HN: What’s a good solution for messy handwriting recognition for mobile app?
Looking for an API, pretrained model or some guidance on how to implement handwriting recognition of post it notes, handwritten paper notes etc. for a mobile app.
- YossarianFrPrez 5y agoThe canonical example of image recognition for AI is recognizing handwritten digits from the NIST dataset; you'll find it in most computer vision tutorials. If you want to learn about rolling your own: https://towardsdatascience.com/classifying-mnist-digits-using-pca-deep-learning-852aeaff48a5 https://towardsdatascience.com/classifying-mnist-digits-usin... https://machinelearningmastery.com/handwritten-digit-recognition-using-convolutional-neural-networks-python-keras/ https://machinelearningmastery.com/handwritten-digit-recogni... https://scikit-learn.org/stable/auto_examples/classification/plot_digits_classification.html https://scikit-learn.org/stable/auto_examples/classification... The only issue is that you'll have to find training data for letters. I'd be surprised if that didn't exist. Also, Google has an API for recognizing handwriting: https://cloud.google.com/vision/docs/handwriting https://cloud.google.com/vision/docs/handwriting I haven't used it, so I can't vouch for it one way or another. But that'd be a pretty easy way to go!
- faizshah 5y agoI’m looking for something specific to messy handwriting, the kind you find in notebooks and on post it notes.
- faizshah 5y agoI found an interesting solution from caltech: https://github.com/caltechlibrary/handprint https://github.com/caltechlibrary/handprint Also it seems the keyword to find models and apis for this is “handwritten text recognition.”