How to annotate complete images?

Question

I am currently playing around with tensorflows object detection to learn the basics.

Now I've set myself the goal to detect letters in computer written text. For example the header of a newspaper article.

I know that object detection might not be the way to go for letter detection but I wanted to know how well a object detection model performes when input data is perfectly similar( computer generated fonts).

My Question:

I encountered the problem that manually annotating each letter in given headers is a really fiddly job and feels not good for training a model.

Now I ask myself this: I could write an Adobe script to generate me an image for every letter of a font(or multiple fonts) with different backgrounds and rotation and save it to my filesystem. -> this would result in (10 Fonts * 256 characters * 10 backgrounds * angles) 256000 images for model training.

For training purposes this is a great amount of data but if I want to annotate each image myself it will get nifty especially because each 1000 images have the same tag..

Question 1:

Is there a package or tool which annotates the complete image with a given tag?

Question 2:

Will training a object detection model with "cropped images" even work or is a rational approach?

Every help/suggestion is highly appreciated! Thanks in advance :)

Answer 1

Is there a package or tool which annotates the complete image with a given tag?

For your use case you don't need to detect each letter. you can try to detect header and annotate the bbox using tools like labelimg and doccano

The bbox for each tagged image will be saved in the format you want and utilised by model.