3 ms·
It's more like a "learnèd machine" approach. No need to train the model, but a pre-trained model may be (or not be, depending on circumstances!) more efficient
by nine_k 7y ago
It's more like a "learnèd machine" approach. No need to train the model, but a pre-trained model may be (or not be, depending on circumstances!) more efficient than "hand-written" OCR approaches.
- jlarocco 7y agoThe article mentions using a bunch of GPUs in AWS to do the processing, so I'm guessing that's a pretty firm "no" in this case...