Original image text alignment extraction through deep learning

Job ID: 36733859

Budget: $30 – $250 USD

First, you need to consider the overall visual features of the image. Deep learning should be used to consider various visual features.

Each text block should be defined as a flashification problem, and classification should be carried out for each paragraph.

In order to train a deep learning model, data labeling must be done first.
To do so, first of all, a person must manually label to check the alignment.

When a deep learning model receives an entire image as input, it needs to obtain some high-level features through a backbone network.

Also, since we have information about the position of the text box before, we can isolate only the features that fit each position from the extracted high-level features.

Each feature separated in this way must be made into a vect

Each vector thus generated represents a probability value of the alignment. The one with the highest probability indicates the alignment of the block.


A solution for image translation. During the process of translating foreign language text in the image, the text alignment between the original image and the output image is not correct, so we are looking for an expert to solve the problem.
Related categories: Python OpenCV Deep Learning