Template Credit: Adapted from a template made available by Dr. Jason Brownlee of Machine Learning Mastery.
SUMMARY: This project aims to construct a predictive model using a TensorFlow convolutional neural network (CNN) and document the end-to-end steps using a template. The Chinese MNIST Characters dataset is a multi-class classification situation where we attempt to predict one of several (more than two) possible outcomes.
INTRODUCTION: The Chinese MNIST dataset uses data collected by Dr. K Nazarpour and Dr. M Chen for a project at Newcastle University. One hundred Chinese nationals took part in data collection. Each participant wrote with a standard black ink pen all 15 numbers in a table with 15 designated regions drawn on a white A4 paper. Each participant repeated this process ten times, with each sheet scanned at the 300×300 pixels resolution. It resulted in a dataset of 15000 images, each representing one character from a set of 15 characters.
In this iteration, we will construct a few simple CNN models to predict the shoe category based on the available images.
ANALYSIS: The one-layer CNN model’s performance achieved an average accuracy score of 92.79% on the test dataset after 15 epochs. The three-layer CNN model processed the same test dataset with an average accuracy measurement of 97.92%.
CONCLUSION: In this iteration, the simple CNN models appeared to be suitable for modeling this dataset. We should consider experimenting with TensorFlow for further modeling.
Dataset Used: Chinese MNIST Digit Recognizer
Dataset ML Model: Multi-class image classification with numerical attributes
Dataset Reference: https://www.kaggle.com/fedesoriano/chinese-mnist-digit-recognizer
One potential source of performance benchmarks: https://data.ncl.ac.uk/articles/Handwritten_Chinese_Numbers/10280831/1
The HTML formatted report can be found here on GitHub.