How vishalla/mathai-prototype works
Project type and purpose
• From the file tree and README ("MathAI"), this is a desktop‐based, PyQt-driven whiteboard application with an integrated convolutional neural network (CNN) for math expression recognition. The system consists of a front-end drawing UI, a networked server component, and a CNN inference pipeline.
File structure
• Top-level files:
• app.py, main.py: entry points for the application
• requirements.txt: Python dependencies (likely PyQt5, PyTorch, NumPy, etc.)
• weights.txt and model_parameters_for_CNN.pth: serialized model weights
• CNN_Model directory:
• Covonutional_neural_network/: core CNN training/inference code (CNNnetwork.py, model_pipeline.py) and utilities (augment_dataset.py, loaddataset.py, model_utils.py)
• Utils/: preprocessing, contour detection, prediction wrappers
• IPython notebooks: exploratory analysis (confusion_matrix.ipynb, pca.ipynb)
• Whiteboard directory:
• UI/: boardscene.py, mainwindow.py, board.ui form definitions
• Server/: TCPServer sub-module handling network sockets and invoking predictions, plus helpers (getip.py, netManage.py, tcpServerNet.py)
• board.py: client-side canvas logic
README for additional insights
• The README is minimal; project name “MathAI” implies a focus on mathematical handwriting recognition. All other context comes from directory naming and file content.