AUTOMATED STUDENT ASSESSMENT: LABEL RECOGNITION IN STEM FIGURES
Visual representations such as Free body diagrams are an important part of solving engineering mechanics problems. Automatic Assessment of these types of images is difficult due to the involvement of multiple object types and to their contextual nature. Using a probabilistic approach, an algorithm was created to automatically categorize groups of characters in labels from images into specific object types including: variables, assignment operators, values, units, or words. Using these categories, the algorithm was then able to determine whether the label was an identifier, a point, a dimension, a variable definition, or an equation. A series of representative test cases were chosen and results found that the current algorithm was able to correctly predict the results of all test cases. The paper discusses each step in detail and provides the resulting probability coefficients for the model.