IV. Results
To achieve the objective of this research, a methodology involving
pre-processing the uploaded flowchart image, text extraction from the
flowchart image using OCR, code generation with large language model
inference was implemented. A diverse test dataset of flowchart images
representing various programming tasks was created. This dataset was
used for evaluating the created application. Python codes generated for
the test flowchart images were executed in a code editor and the outputs
produced were checked. Quality of the Python codes generated for the
test flowchart images was evaluated for their accuracy, syntactic
correctness, and adherence to best programming practices. Codes
generated for 75% of the flowchart images from the test dataset
executed without any compile time and runtime errors. Best programming
practices were followed in 64% of the generated python codes for test
flowcharts. The experiments demonstrated promising results, with the
application successfully generating Python code and its brief
explanation from flowchart images with a high degree of accuracy. It was
observed that the performance of the application varied depending on the
complexity of the flowchart and the clarity of the flowchart images.
However, even for 62% of challenging cases where uploaded flowchart
images were not clear and the text in the flowchart components wasn’t
clearly visible, The application was able to generate the correct python
codes.