This project counts grid cells from spreadsheet-like images, even with slight variations in angles (~5 to 10 degrees). Will be extended to perform OCR on extracted text.
- Python 3.6.3 or higher
- numPy 1.16.0 or higher:
pip install numpy - OpenCV 3.4.0 or higher:
pip install opencv-python - pandas 0.20.3 or higher:
pip install pandas - matplotlib 2.1.0 or higher:
pip install matplotlib
- soln.py: Main source file
- docs.md: Algorithmic approach and possible extensions
- data_augment.py: Scipt to generate artificially rotated images (suspended for now)
Run the main file python soln.py from the project directory. Make sure to rename/delete the existing output folder if you want to obtain your own results. Any image added to the input directory will automatically be considered for processing.
Apart from the results being printed out, the generated output.csv file holds the results of the number of grid cells for each image in the input folder. The output folder also holds images of contours drawn over the original image along with histograms to fine-tune threshold values used in creation of intermediate binary image.