Manually segment representative images

[d-genk]

For each representative image, create a structured representation of pixel coordinates for all discrete regions of entry text in the image at its original aspect ratio. Do the same for each image when rescaled to 960 x 1280. This could either also be done manually or by the application of a transformation algorithm to the original coordinate set. The final dataset should be a JSON document containing a record for each image with both sets of coordinates.

[d-genk]

Annotation tool: https://recogito.pelagios.org/