How do I add cursive ocr capabilities? #87
-
|
Hi, sorry, the whole concept of training and adding models are very new to me. As far as I know, the program works for extracting texts from images. When I tried working with images with cursive handwriting, - it doesn't work. What are the procedures I have to do to add cursive extraction capabilities? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
|
You would need to train, or more likely fine-tune, the recognition model (and maybe the detection one) on a dataset of handwriting images. The ocrs-models repository has notes on how training of the current models was done. The documentation assumes some experience doing ML model training in PyTorch. |
Beta Was this translation helpful? Give feedback.
-
|
I see, this is informative thank you! |
Beta Was this translation helpful? Give feedback.
-
|
For ML training the alternative to owning a GPU is to rent one in the cloud, by the hour, from a service such as RunPod or AWS. Their cheapest options will be good enough. You can run the training script on a CPU, and this is useful for testing that it works, but it will be very slow. What I usually do is to do a small test run on a sliver of the training data locally on the CPU to check the code runs, then do the actual training in the cloud on a GPU. |
Beta Was this translation helpful? Give feedback.
-
|
I see, thank you for the information! It's much appreciated. |
Beta Was this translation helpful? Give feedback.
You would need to train, or more likely fine-tune, the recognition model (and maybe the detection one) on a dataset of handwriting images. The ocrs-models repository has notes on how training of the current models was done. The documentation assumes some experience doing ML model training in PyTorch.