This project has been created for the May 2023 IB Computer Science IA Examination.
Hosted at: mathconverter.pythonanywhere.com
The functions in this file assume you have cv2 and numpy installed.
The get_contours function takes in an image path as a string and outputs an array of contours that can be used with the cv2.boundingRect() function to get bounding boxes around key features in an image.
The contrast can be changed by adjusting the value inside the brackets on line 10. The lower the value, the higher the contrast, however it can cut into the image causing it to not be detected properly.
Example:
contours = get_contours("path/to/image/here")The sort_chars function takes in an array of contours as input and outputs the same array of contours that is sorted in reading order (left-to-right).
This function uses bubble sort to sort the array as there would be very few items to sort.
Example:
contours = get_contours("path/to/image/here")
sorted_contours = sort_chars(contours)The get_chars function takes in an image path as a string, an array of contours, an integer defining the area threshold, an integer defining the border size, and boolean of whether debugging tools are needed.
The contours can be taken from the get_contours function.
The area threshold determines the minimum allowed area for the bounding box that can be accepted and used. This value can be any integer more than 0.
The border size parameter defines the thickness of the added border in pixels for better OCR. This value can be any non-negative integer.
The debug mode displays the image that has been inputted and draws a bounding box around all allowed areas. This value is optional and is False by default.
Example:
image_path = "path/to/image/here"
contours = get_contours(image_path)
characters = get_chars(image_path, contours, 750, 30, False)The functions in this file assume os, torch, torchvision, PIL, and sympy are already installed.
This also requires another file called neural_network.py with a class called NeuralNetwork to be in the same directory as this one.
The predict function takes in an array of characters, a string referencing the model path, and a boolean stating whether debug tools are needed. It outputs a string of all the predictions combined into one.
The array of characters can be taken from the get_chars function.
The model path is a string pointing to the model that would be used. This is depedent on the neural_network.py file.
The debug parameter displays each of the predictions that the network outputs. This is an optional setting and is set to False by default.
Example:
image_path = "path/to/image/here"
model_path = "path/to/model/here"
contours = get_contours(image_path)
characters = get_chars(image_path, contours, 750, 30, False)
predicted_equation = predict(characters, model_path, False)The combine_equation function takes an equation as a string and outputs a string to be used for rendering and another for solving.
The equation can be inputted manually or can also be taken from the predict function.
The rendering equation is a string that is formatted for LaTeX rendering (can use MathJax to render LaTeX in webpages).
The solving equation is a string that is formatted for use in the solver function.
Example:
render_equation, solve_equation = combine_equation("2x+6=12")The solver equation takes in an equation as a string and a boolean for if debug tools are needed, and outputs an answer as a string.
The inputted equation can be taken from the solving equation of the combine_equation function.
The debug tools are optional and are False by default.
The output is a string which has been LaTeX formatted for displaying on a webpage through the use of MathJax.
Examples:
render_equation, solve_equation = combine_equation("2x+6=12")
rendered_answer = solver(solve_equation)The render_math function is an internally used function to convert the answer into a LaTeX formatted version which is then returned.
It takes an string as an input and formats it to follow the LaTeX rules using the SymPy library.
Example:
rendered_answer = render_math("3/4 + 27")Usage of all the functions listed above can be seen in the example_usage.py file.