Handwritten Digit Recognition
Personal portfolio python project that will be implementing a machine learning classification model to predict the value of a handwritten digit through supervised learning.
About
Reading letters & numbers off of a sheet of paper is profoundly intuitive for us humans, as our visual cortex has developed and evolved for thousands of years. For computers, on the other hand, interpreting strokes from a sheet of paper algorithmically is incredibly difficult. However, due to their immense processing speed, computers can take in massive amounts of data to therefore effectively learn and predict what a particular digit means.
The model that I've opted for this particular project is the SVM model, or linear Support Vector classification. Although there are different approaches it is difficult to appoint a specific model as superior, as it depends on the criteria or project scope. However, SVM is the preferred method as it allows for the fine-tuning of parameters to achieve better accuracy.
Dependencies
The algorithm itself will be given a large dataset, MNIST, from sci-kit learn. Tools include: SciKit-learn, matplotlib
Versions
Low-Level In the lower-level approach, you will see how a computer can read a large data set and output the users chosen element from the data structure, in a non-robust form.
High-Level (work in progress) In the latter, higher-level, the user will be able to simply write a digit, through an input dialog, for the computer to interpret, using additional web components.