# Purpose: Improve Keras-OCR to read detected texts in a more 'human' way, meaning from left to right and top to bottom. # Install all necessary libraries (Note: I used a conda environment.) pip install -r requirements.txt Full Explanation: https://shegocodes.medium.com/extract-text-from-image-left-to-right-and-top-to-bottom-with-keras-ocr-b56f098a6efe Sample Runs: # Raw detection (order does not matter) python detect.py --image images/wanderer.png --thresh 10 --order no # Tighten threshold python detect.py --image images/wanderer.png --thresh 10 --order yes # Average threshold python detect.py --image images/wanderer.png --thresh 15 --order yes # Loosen threshold python detect.py --image images/wanderer.png --thresh 20 --order yes
shegocodes/keras-ocr
Extract Text from Image Left-to-Right and Top-to-Bottom with Keras-OCR
Jupyter Notebook