This machine learning project aims to recognize handwritten mathematical symbols using the HASYv2 dataset. The dataset consists of hand-drawn mathematical symbols, and our goal is to classify these symbols into their respective categories. The project is divided into two parts: Part 1 and Part 2, each using different machine learning techniques.
The HASYv2 dataset is a collection of hand-drawn mathematical symbols. It includes various symbols used in mathematics and engineering, such as numbers, operators, and special symbols. The dataset is organized into training and test sets.
-
Clone this repository to your local machine:
git clone https://github.com/baudoin-csp/math-symbol-recognition.git -
Navigate to the project directory:
cd math-symbol-recognition
-
To run K-Means clustering (e.g., with K=3), use the following command:
python part1/main.py --data data/hasy-data --method kmeans --K 3 -
To run Logistic Regression (e.g., with lr=0.00001 and max_iters=100), use the following command:
python part1/main.py --data data/hasy-data --method logistic_regression --lr 1e-5 --max_iters 100 -
To run SVM (e.g., with C=1, an RBF kernel, and gamma=0.01), use the following command:
python part1/main.py --data data/hasy-data --method svm --svm_c 1. --svm_kernel rbf --svm_gamma 0.01
-
To run the MLP model in Part 2, use the following command:
python part2/main.py --data data/hasy-data --method mlp -
To run the CNN model in Part 2, use the following command:
python part2/main.py --data data/hasy-data --method cnn
In both parts, you can add the --test argument to predict on the actual test set; otherwise, your code should use a validation set.
The project reports for both Part 1 and Part 2 are available in the their respective directory as PDF files. These reports provide detailed information about our methodology and a summary of the results achieved using different methods.
This project is licensed under the MIT License - see the LICENSE file for details.
We welcome feedback, bug reports, and contributions. Please feel free to create issues or pull requests if you have any suggestions or improvements to make.