How dbetm/handwritten-flowchart-with-cnn works
File Structure Analysis
• Top-level orchestration:
• handler.py – CLI interface
• preprocessor.py, segmenter.py, splitter.py – pipeline stages to clean, split and segment the input image
• code_generator.py – translates the intermediate representation into C code
• flowchart_generator/ – module that uses Graphviz to build .dot files and render flowchart images
• Model directories:
• text_model – text detection: data/ for generator/preprocessing, network/ for CNN+LSTM model, output/ for checkpoints and logs
• Results and assets:
• results/ – holds generated C code, compiled object files, and flowchart outputs
• Images/ and graph_images/ – sample inputs and intermediate visuals
• Config and dependencies:
• requirements.txt / requirements-dev.txt
• model/args.json and various fix_input_weights_path.py scripts for path corrections
Key README Insights
• Use of continual learning to refine the text model per user
• Final outputs: C source code, compiled binaries, digital flowchart image