Grid/table hand written numbers

[octaviordz]

Is a valid use case the processing of images such as this one [1]?, in a simple execution I was not able to get 'any' useful data from running ocrs with the given image.

[1]

[robertknight]

Ocrs's current model is mostly trained on photos of printed text (see dataset examples) so it will struggle with an image like this.

One way to approach this would be to use classical computer vision techniques (example) to find the rows and columns and then extract images of individual cells or rows which will be easier for a model to process. It will likely be easier to use Python to do some prototyping and port the result to Rust once you have something that works.